Publication number | US7409415 B2 |

Publication type | Grant |

Application number | US 10/326,515 |

Publication date | Aug 5, 2008 |

Filing date | Dec 20, 2002 |

Priority date | Dec 20, 2002 |

Fee status | Paid |

Also published as | EP1431870A2, EP1431870A3, US20040123079 |

Publication number | 10326515, 326515, US 7409415 B2, US 7409415B2, US-B2-7409415, US7409415 B2, US7409415B2 |

Inventors | Patrick W. Bosshart |

Original Assignee | Texas Instruments Incorporated |

Export Citation | BiBTeX, EndNote, RefMan |

Patent Citations (5), Non-Patent Citations (2), Referenced by (8), Classifications (10), Legal Events (3) | |

External Links: USPTO, USPTO Assignment, Espacenet | |

US 7409415 B2

Abstract

An electronic system (**200** _{1}) for manipulating an input data argument (D[31:0]) comprising an integer number of bits. The system comprises an input (R) for receiving a right direction argument and an input (L) for receiving a left direction argument. The system also comprises circuitry (**200**) for producing a first data output having the integer number of bits by rotating the input data argument in response to the first direction argument and the second direction argument. The system also comprises circuitry for providing a modified data output (**502**). The circuitry for providing comprises circuitry for selecting a first set of bits from the first data output as a first portion of the modified data output and circuitry for providing a second set of bits from a source other than the first data output as a second portion of the modified data output.

Claims(39)

1. An electronic system for manipulating an input data argument comprising an integer number of bits, the system comprising:

an input for receiving a right direction argument;

an input for receiving a left direction argument;

circuitry for producing a first data output having the integer number of bits by rotating the input data argument in response to the right direction argument and the left direction argument in a single shift operation; and

circuitry for providing a modified data output, comprising:

circuitry for selecting a first set of bits from the first data output as a first portion of the modified data output; and

circuitry for providing a second set of bits from a source other than the first data output as a second portion of the modified data output.

2. The electronic system of claim 1 wherein the circuitry for producing a first data output having the integer number of bits by rotating the input data argument and the circuitry for providing a modified data output both operate in response to an electronic instruction.

3. The electronic system of claim 2 wherein the electronic instruction is selected from a set consisting of EXTRACT, SHIFT RIGHT, SHIFT LEFT, SET, CLEAR, and SATURATING SHIFT LEFT.

4. The electronic system of claim 2 :

wherein the instruction comprises a FUNNEL SHIFT instruction; and

wherein the circuitry for providing a second set of bits is disabled for the FUNNEL SHIFT instruction.

5. The electronic system of claim 1 wherein the circuitry for providing a second set of bits provides a sign bit for each bit of the second set of bits.

6. The electronic system of claim 5 :

wherein the data argument comprises the integer number of bit positions;

wherein each of the bit positions corresponds to a respective one of the integer number of bits; and

wherein the circuitry for providing provides the sign bit as a bit from the input data argument located in the input data at a bit position equal to the integer number of bits minus the left direction argument.

7. The electronic system of claim 6 and further comprising circuitry for generating a code, wherein the circuitry for selecting selects each bit in the first set of bits in response to each bit in the code that has a first state and wherein the circuitry for providing provides each bit in the second set of bits in response to each bit in the code that has a second state different than the first state.

8. The electronic system of claim 7 wherein the code comprises a thermometer code.

9. The electronic system of claim 8 :

wherein the thermometer code comprises of a first number of bits that each has the first state; and

wherein the first number equals the integer number minus the right direction argument.

10. The electronic system of claim 9 :

wherein the thermometer code comprises a second number of bits that each has the second state; and

wherein the second number equals the right direction argument.

11. The electronic system of claim 9 and further comprising, responsive to the right direction argument being larger than the left direction argument, circuitry for providing a final output consisting of the modified output.

12. The electronic system of claim 11 wherein the circuitry for producing a first data output and the circuitry for providing a modified output both operate in response to an electronic EXTRACT instruction.

13. The electronic system of claim 9 and further comprising, responsive to the left direction argument being larger than the right direction argument, circuitry for providing a final output comprising a portion of the modified output.

14. The electronic system of claim 13 wherein the circuitry for providing the final output comprises:

circuitry for including in the final output a value of zero in a number of least significant bit positions of the final output, wherein the first number of least significant bit positions equals the left direction argument minus the right direction argument; and

circuitry for including in the final output a value from a corresponding bit position in the modified output in a number of most significant bit positions, wherein the number of most significant bit positions equals the integer number minus the number of least significant bit positions.

15. The electronic system of claim 14 wherein the circuitry for producing a first data output and the circuitry for providing a modified output both operate in response to an electronic EXTRACT instruction.

16. The electronic system of claim 1 wherein the circuitry for providing a second set of provides selects a value of zero for each bit of the second set of bits.

17. The electronic system of claim 16 :

wherein the second set of bits consists of a number of least significant bits of the modified output; and

wherein the number of least significant bits equals the left direction argument.

18. The electronic system of claim 17 wherein the circuitry for providing a modified data output operates in response to an electronic SHIFT LEFT instruction.

19. The electronic system of claim 16 and further comprising circuitry for generating a final thermometer code, wherein the second set of bits are provided in response to bits in the final thermometer code that have a first state, and wherein the circuitry for generating the final thermometer code comprises:

circuitry for forming a complement of the left direction argument;

circuitry for forming an initial thermometer code having a first number of bits in the first state and a second number of bits in the second state, wherein the first number equals the complement of the left direction argument and the second number is equal to the integer number minus the first number;

circuitry for forming a complement of the initial thermometer code; and

circuitry for shifting the complement of the initial thermometer code rightward one bit position to form the final thermometer code.

20. The electronic system of claim 19 wherein the circuitry for providing a modified data output operates in response to an electronic SHIFT LEFT instruction.

21. The electronic system of claim 1 :

wherein the source other than the first data output comprises a first source;

wherein the circuitry for selecting a first set of bits is further operable to select the first set of bits from a second source other than the first data output, wherein the second source provides a fixed data state; and

wherein the circuitry for providing a second set of bits from the first source is operable to provide the second set of bits from the input data argument.

22. The electronic system of claim 21 wherein the circuitry for providing a modified data output operates in response to an electronic instruction selected from a set consisting of a SET and a CLEAR instruction.

23. The electronic system of claim 22 :

wherein the electronic instruction consists of a SET instruction; and

wherein the fixed data state consists of a binary one.

24. The electronic system of claim 22 :

wherein the electronic instruction consists of a SET instruction; and

wherein the fixed data state consists of a binary zero.

25. The electronic system of claim 21 wherein the circuitry for selecting the first set of bits is operable to select, for the modified data output, the first set of bits as a block of bits positioned, between and including, a first bit position identified in response to the right direction argument and a second bit position identified in response to the left direction argument.

26. The electronic system of claim 25 wherein the circuitry for providing the second set of bits is operable to provide each bit for the modified data output, other than the block of bits, as a bit from a respective bit position in the input data argument.

27. The electronic system of claim 26 wherein the circuitry for providing a modified data output operates in response to an electronic instruction selected from a set consisting of a SET and a CLEAR instruction.

28. The electronic system of claim 1 and further comprising circuitry for providing the input data argument by selecting a plurality of bytes from a first data quantity having a number of bytes and from a second data quantity having the number of bytes, wherein the plurality of bytes comprises at least one byte from each of the first data quantity and the second data quantity.

29. The electronic system of claim 28 :

wherein the circuitry for providing the input data and the circuitry for providing the modified data output both operate in response to an electronic FUNNEL SHIFT instruction; and

wherein the circuitry for providing a second set of bits is disabled for the FUNNEL SHIFT instruction.

30. The electronic system of claim 1 and further comprising:

circuitry for detecting saturation that can result in response to the left direction argument and the input data argument; and

circuitry, responsive to the circuitry for detecting saturation, for outputting a final output comprising a saturation constant in place of the modified data output.

31. The electronic system of claim 30 wherein the circuitry for outputting the final output is operable to create the final output by logically combining the modified data output with a plurality of logic signals.

32. The electronic system of claim 30 wherein the saturation constant is selected from a set consisting of a positive saturation constant and a negative saturation constant.

33. The electronic system of claim 1 wherein the circuitry for providing a modified data output comprises static logic circuitry.

34. The electronic system of claim 1 wherein the circuitry for providing a modified data output comprises dynamic logic circuitry.

35. The electronic system of claim 1 wherein the modified data output consists of the first set of bits and the second set of bits.

36. The electronic system of claim 35 wherein the modified data output consists of 32 bits.

37. The electronic system of claim 1 wherein the modified data output consists of 32 bits.

38. The electronic system of claim 1 wherein the circuitry for producing and the circuitry for providing a modified data output are part of a processor.

39. The electronic system of claim 1 wherein the circuitry for producing and the circuitry for providing a modified data output are part of a digital signal processor.

Description

This application is related to U.S. patent application Ser. No. 10/326,704, entitled “Processor System And Method With Combined Data Left and Right Shift Operation”, and filed on the same date as the present application.

Not Applicable.

The present embodiments relate to electronic processors such as microprocessors or digital signal processors (“DSPs”), and are more particularly directed to providing such a processor operable in response to an EXTRACT instruction to efficiently shift a data argument in response to respective left and right shift count indicators.

Electronic processors including microprocessors and DSPs have become prevalent in numerous applications, including uses for devices in personal, business, and other environments. Demands of the marketplace affect many aspects of processor design, including factors such as device power consumption and speed. As a result, constant evaluation is performed by processor designers in an effort to provide improvements in these and other factors. The present embodiments also endeavor toward this end.

The present embodiments are directed to providing an improved processor functional unit for purposes of rotating a data argument. Rotation of a data argument is intended to cover the instance where data is to be shifted twice, once in a first direction and then thereafter in an opposite direction. Thus, two possibilities exist, namely, a right shift followed by a left shift or a left shift followed by a right shift. Two such successive shifts may be required in various instances, where by way of example the present commercially available DSPs sold by Texas Instruments Incorporated under the TMS320C6 xxx family include an EXTRACT instruction in the DSP instruction set, and that instruction calls for two successive and opposite-direction shifts. The use of such operations is typically beneficial to manipulate or isolate a portion of data within a data word, where the word in contemporary processors is often a 32-bit data argument. Further, these two shifts can be used with additional operations, such as arithmetic or logical operations with another data argument, in order to achieve an intended result. In any event, in previous processors, such operations required either the use of two successive shift instructions, or the use of the EXTRACT instruction, with either approach being executed in the processor by performing a complete shift in one direction followed by another complete shift in the opposite direction. In this regard, either two independent operations are expended, one for each shift, or the two shifts might be executed simultaneously, thereby requiring however the complexity of two different 32-bit shifters (i.e., one for each of the two shifts). Additional delay and/or additional complexity, such as that required by these prior art approaches, are considered drawbacks in contemporary processor design.

In view of the above, there arises a need to address the drawbacks of the prior art, as is achieved by the preferred embodiments described below.

In the preferred embodiment, there is an electronic system for manipulating an input data argument comprising an integer number of bits. The system comprises an input for receiving a right direction argument and an input for receiving a left direction argument. The system also comprises circuitry for producing a first data output having the integer number of bits by rotating the input data argument in response to the first direction argument and the second direction argument. The system also comprises circuitry for providing a modified data output. The circuitry for providing comprises circuitry for selecting a first set of bits from the first data output as a first portion of the modified data output and circuitry for providing a second set of bits from a source other than the first data output as a second portion of the modified data output.

Other aspects are also disclosed and claimed.

_{1}, and to demonstrate a shift example the shift arguments are R=10 and L=6 such that right shift R is greater than the left shift L.

_{2}, and to demonstrate a shift example the shift arguments are R=8 and L=16 such that right shift R is less than the left shift L.

_{ARG1 }that represents the actual physical pins that are input to data rotate unit **200** of

**600** for implementing various of the signal selectivity provided by multiplexer **502**, saturation constant switch **582**, and AND gate **504** of _{EXEC} _{ — } _{F}.

**590** of

**600** _{DL }for implementing various of the signal selectivity provided by multiplexer **502**, saturation constant switch **582**, and AND gate **504** of _{EXEC} _{ — } _{F }and V_{EXEC} _{ — } _{FX}.

By way of introduction to the preferred embodiment processor and its operability to shift a data argument in response to separate left and right shift count indicators,

_{1}, where for sake of example its bits are designated D31 through D0. In the example of _{1 }is shifted left 6 bit positions, thereby discarding bits D26 through D31 with the left shift and causing the shifted value VL_{1 }to begin with and include the remaining 26 bits of the 32 bits of V_{1}; further, the remaining 6 least significant bits (“LSBs”) of VL_{1 }are filled with zeros. The right shift is next performed, during which the 32-bit value VL_{1 }is right shifted 10 bit positions, thereby filling the 10 most significant bits (“MSBs”)of the right shifted value, VR_{1}, with zeros; alternatively, these bits may each be a sign extension bit, as abbreviated in _{1}, is the 22 most significant bits from value VL_{1}, that is, bits D25 through D4.

_{2}, using the _{2 }is shifted left 16 bit positions, thereby discarding bits D31 through D16 so that the left shifted value VL_{2 }begins with and includes the remaining 16 bits of the 32 bits of V_{2}, and the remaining 8 LSBs are filled with zeros. The right shift is next performed, during which the 32-bit value VL_{2 }is right shifted 8 bit positions, thereby filling the 8 MSBs of the right shifted value, VR_{2}, with zeros (or alternatively, sign extension bits). Following the 8 shifted-in zeros (or sign extension bits), and thereby completing the 32-bit value VR_{2}, is the 24 MSBs from value VR_{1}; however, of those 24 bits, only the 16 data bits D15 through D0 remain from the original value V_{2}, where the remaining 8 LSBs are values of zero.

**100**, where processor **100** may be a microprocessor or DSP or still other computing integrated circuit. Processor **100** includes numerous aspects known in the art and, to focus on aspects of the preferred embodiment, such known aspects are neither shown nor described. Generally, however, one skilled in the art will recognize that processor **100** will include one or more functional units, where each such functional unit typically includes some type of sub-division of functionality and may be referred to herein as a sub-unit. As known in the art, processor **100** has an associated instruction set, and each sub-unit operates to perform data operations with respect to one or more data arguments according to any one of the instructions in that instruction set. Such sub-units may be, by way of example, an arithmetic logic unit (“ALU”), a logical operator block that is operable to perform various logical operations, such as are known in the art to include by ways of example logical AND, OR, and so forth, as well as others. Further with reference to the sub-unit functionality, **200** that is directed to the left/right shifting introduced above, where sub-unit **200** may exist separately from other sub-units or may be combined with the functionality of one of those sub-units. Thus, sub-unit **200** may be considered to be a part of a shifter unit. The specific terminology is not critical, where one skilled in the art will appreciate from the remainder of this document that sub-unit **200** performs what will be referred to as a rotation operation with respect to both the left and right shift arguments and, thus, sub-unit **200** may be considered a data rotate unit **200**.

Additional introduction is now provided with respect to the name “data rotate unit” in that the unit rotates a 32-bit data input in response to a left and right shift argument L and R, respectively. To further understand this aspect, attention is returned to _{1 }reflects a net shift, or a rotation, of R−L=10−6=4, that is, a shift to the right of 4 data bits, as shown in that the value VR_{1 }does not include the 4 data bits D3, D2, D1 and D0 of V_{1}. Further, the first data bit, D25, in the final value VR_{1}, as retained from the original value of V_{1}, is the most significant bit retained following the left shift of 6 bits (which thereby discards bits D31 through D28). As another example, with reference to _{2 }reflects a net shift, or a rotation, of R−L=8−16=−8, where the negative indication in this result indicates a net shift to the left, and the shift is of the difference of 8 data bits; further, the first data bit, D15, in the final value, VR_{2}, as retained from the original value, V_{2}, is the most significant bit retained following the left shift of 16 bits. Given these two examples, one skilled in the art should appreciate that the left shift followed by the right shift provides, in effect a rotation of the original value V_{x}, where other circuitry, if needed, will insert the most significant bit zeros (or sign extensions) and the lest significant bit values of zero.

Having demonstrated the rotation functionality of two opposite direction shifts, the preferred embodiments achieve such functionality without two separate shifters or two independent and complete shifts which require that the data argument is fully shifted in one direction followed by next shifting it in the opposite direction. Instead, in one embodiment, such an approach may be implemented by providing a control circuit sufficient to determine R−L and to rotate the values accordingly (with sufficient additional most significant bit zeros or sign extensions when appropriate). However, such an approach may be unduly complex and give rise to considerable delay. Therefore, various preferred embodiments implement data rotate unit **200** in view of the notion of the R−L function, but in a way that first expresses that function in a different manner and that also breaks down the function based on subsets of the bits in the data shift arguments R and L. These aspects are further explored in the remainder of this document.

In the preferred embodiment, it is recognized that the net shift of R−L as described above is equally realized as in the following Equation 1:

*R−L=R+LX+*1 Equation 1

In Equation 1, the “X” indication following a variable, as used throughout this document, is an indication of the inverted signal; thus, in this example, LX is the logical inverse (i.e., binary complement) of L. Further, the Equation 1 principle by itself is known in the art of data shifting, that is, that a rotation of R−L equals a rotation of R+LX+1. To further appreciate this principle, consider it as applied to the examples of

Returning now to **200**, it includes various ROTATE STAGEs and connections to determine the value R+LX+1 with respect to a 32-bit data argument D[31:0], where R and L are both 5-bit shift argument values shown as L[4:0] and R[4:0]. Each of these three arguments may be provided from various locations within processor **100**, such as in response to an instruction and as loaded from a register, register file, memory, or the like. Each ROTATE STAGE represents a group of 32 multiplexers that is connected to either pass the 32-bit input straight through, or to right rotate it a number of locations, that is, to shift the input to the right such that any bits shifted out of one end of the stage wrap around to an input of the other end of the stages. Further, the amount of rotation for each stage differs for certain of the various stages as detailed later. Turning now to the specific connections, the data argument D[31:0] is connected to a ROTATE STAGE 1 that is connected to receive the LSB, L[0], of the left shift argument. In the preferred embodiment, it is presumed that the inverse of each shift argument is also available, or one skilled in the art may readily implement sufficient circuitry to provide such inverses, and in any event therefore and as detailed later each stage that receives a shift argument is further connected to receive the inverse of that argument. Thus, the ROTATE STAGE 1, while not explicitly illustrated due to the above-stated assumption, also receives or has access to the inverse of L[0], referred to hereafter as LX[0]. The 32-bit output of the ROTATE STAGE 1 is connected as an input to a ROTATE STAGE 2. The ROTATE STAGE 2 is connected to receive the least significant bit, R[0], of the right shift argument and, thus, as stated above, it is also preferably connected to receive its inverse, RX[0]. Thus, given the connections described thus far, note that the 5-bit shift argument values L and R are connected such that the least significant bit of each provides a control to the ROTATE STAGE 1 or the ROTATE STAGE 2. As detailed later, each of these STAGEs is operable to right rotate its 32-bit input in response to the value of the single bit portion of the shift argument that is also input to the respective STAGE.

Continuing with **200**, the 32-bit output of ROTATE STAGE 2 is connected as an input to ROTATE STAGE 3. ROTATE STAGE 3 also receives a 4-bit control signal, C_{3}, from a STAGE 3 CONTROLLER. The STAGE 3 CONTROLLER is connected to receive the second and third LSBs from each of the left and right shift arguments and, thus, these inputs are indicated in _{4}, to a ROTATE STAGE 4. Lastly, ROTATE STAGE 4 receives at an input the 32-bit output of STAGE 3, and as detailed later, in response to the control signal C_{4}, STAGE 4 is operable to shift this 32-bit input and provide a final 32-bit shifted output signal, V_{S}.

The operation of data rotate unit **200** is now discussed, and is further explored later at a device level after a discussion of the preferred embodiment functionality of the various blocks shown in

Looking in detail to the ROTATE STAGE 1, recall it is controlled solely by bit LX[0]; thus, its rotating, if any, is independent of the right shift argument, R, and also of any of the more significant bits LX[4:1] of the left shift argument, L. The ROTATE STAGE 1 rotates the 32-bit data argument D[31:0] to the right one position if LX[0]=1, and no rotation (i.e., a rotation of 0 bit positions) is performed if LX[0]=0. To illustrate an example of the operation of the ROTATE STAGE 1,

Looking in detail to the ROTATE STAGE 2, recall it is controlled solely by bit R[0]; thus, its rotation, if any, is independent of the left shift argument, L, and also of any of the more significant bits R[4:1] of the right shift argument, R. In response to R[0], the ROTATE STAGE 2 rotates its 32-bit data argument input to the right one position if R[0]=1, and no rotation is performed if R[0]=0. Recall further that the preferred embodiment also effects a rotation of the Equation 1 value of R+LX+1; thus, this equation includes an element of “+1”, that is, a right rotation of one position in addition to the rotation of “R+LX.” In one preferred embodiment, this additional increment is achieved in connection with the ROTATE STAGE 2. Thus, in combination with the value of R[0], the ultimate right rotation of the ROTATE STAGE 2 is such that it rotates its 32-bit data argument input to the right two bit positions if R[0]=1 (i.e., one bit position for the sum of one, and another bit position for the element “+1” in Equation 1), and it rotates its 32-bit data argument input to the right one bit position if R[0]=0 (i.e., only one bit position for the element “+1” in Equation 1). To continue with the example of the operation of the ROTATE STAGE 1 and

Looking in detail to the ROTATE STAGE 3, recall it is controlled by the 4-bit control signal C_{3 }from STAGE 3 CONTROLLER, and the ROTATE STAGE 3 CONTROLLER receives the bits L[2:1] and R[2:1] (and their complements). In connection with accomplishing Equation 1, the ROTATE STAGE 3 CONTROLLER adds LX[2:1] and R[2:1] to form a two-bit sum, with any carry from the MSB location of that sum used to set an asserted state of CARRY to the STAGE 4 CONTROLLER. Thus, the various possibilities generated by the addition of LX[2:1] and R[2:1] are as shown in the following Table 1, along with the resulting state of CARRY and its complement, CARRYX.

TABLE 1 | |||

LX[2:1]; R[2:1] | LX[2:1] + R[2:1] | CARRY | CARRYX |

00; 00 | 00 | 0 | 1 |

00; 01 | 01 | 0 | 1 |

00; 10 | 10 | 0 | 1 |

00; 11 | 11 | 0 | 1 |

01; 00 | 01 | 0 | 1 |

01; 01 | 10 | 0 | 1 |

01; 10 | 11 | 0 | 1 |

01; 11 | 11 | 1 | 0 |

10; 00 | 10 | 0 | 1 |

10; 01 | 11 | 0 | 1 |

10; 10 | 00 | 1 | 0 |

10; 11 | 01 | 1 | 0 |

11; 00 | 11 | 0 | 1 |

11; 01 | 00 | 1 | 0 |

11; 10 | 01 | 1 | 0 |

11; 11 | 10 | 1 | 0 |

In addition, the two bits of the sum generated by the STAGE 3 CONTROLLER determine the state of control signal C

TABLE 2 | ||

LX[2:1]; R[2:1] | LX[2:1] + R[2:1] | C_{3} |

00; 00 | 00 | 0001 |

00; 01 | 01 | 0010 |

00; 10 | 10 | 0100 |

00; 11 | 11 | 1000 |

01; 00 | 01 | 0010 |

01; 01 | 10 | 0100 |

01; 10 | 11 | 1000 |

01; 11 | 00 | 0001 |

10; 00 | 10 | 0100 |

10; 01 | 11 | 1000 |

10; 10 | 00 | 0001 |

10; 11 | 01 | 0010 |

11; 00 | 11 | 1000 |

11; 01 | 00 | 0001 |

11; 10 | 01 | 0010 |

11; 11 | 10 | 0100 |

Lastly, because ROTATE STAGE 3 responds to the second and third least significant bit positions, note that each sum in Table 2 corresponds to a factor times two, that is, each sum 00, 01, 10, and 11 in a binary sense corresponds to a decimal value of 0, 2, 4, and 6, respectively. Accordingly, in response to the value of C

TABLE 3 | ||

LX[2:1] + R[2:1] | C_{3} |
STAGE 3 rotation |

00 | 0001 | 0 positions |

01 | 0010 | 2 positions |

10 | 0100 | 4 positions |

11 | 1000 | 6 positions |

To further appreciate the preceding discussion of the operation of the ROTATE STAGE 3, _{3}=0010 and from Table 3, that value of C_{3 }gives rise to a right rotation of 2 bit positions. Accordingly,

Looking in detail to ROTATE STAGE 4, recall it is controlled by the 4-bit control signal C_{4 }from the STAGE 4 CONTROLLER, and the STAGE 4 CONTROLLER receives both bits L[4:3] and R[4:3] (and their complements). In connection with accomplishing Equation 1, the STAGE 4 CONTROLLER adds LX[4:3] and R[4:31], and also accounts for any asserted CARRY from the STAGE 3 CONTROLLER, to form a two-bit sum. Mathematically the carry can be added to the sum of LX[4:3] and R[4:3], but as detailed later, in the preferred embodiment to increase speed the carry is instead provided to change the decoded value of C_{4 }that results solely from the sum of LX[4:3] and R[4:3]. Looking first to the sum of LX[4:3] and R[4:3], and temporarily not considering the CARRY from the STAGE 3 CONTROLLER, first note that any carry from the MSB location of that sum of LX[4:3] and R[4:3] may be disregarded, as it represents a set bit in the sixth bit position of a binary number, thereby corresponding to a shift of 32 positions; however, since the preferred embodiment is directed to a 32-bit data argument, then a shift of 32 positions simply would return the same value as the input, that is, the set-up is of a modulo 32 structure. Accordingly, there is no need to retain or respond to a carry bit in this sixth bit position. Thus, the two-bits of the sum generated by the STAGE 4 CONTROLLER determine the state of control signal C_{4}, where like the STAGE 3 CONTROLLER here again the sum is decoded such that only one of the four bits in C_{4 }is asserted, with a different bit asserted for each of the four different sums possible for LX[4:3]+R[4:3]. The table for these combinations is the same as that shown above in Table 2, but here as applied to the most significant bit positions (i.e., the fourth and fifth bits) of the shift arguments L and R. However, because the ROTATE STAGE 4 responds to the fourth and fifth significant bit positions, then each sum in Table 2, as applied to the ROTATE STAGE 4, corresponds to a factor times eight, that is, each sum 00, 01, 10, 11 in a binary sense corresponds to a decimal value of 0, 8, 16, and 24, respectively. Accordingly, in response to the value of C_{4}, the ROTATE STAGE 4 shifts its 32-bit input data argument a bit distance of either 0, 8, 16, or 24 positions, depending on the respective indication of C_{4}, and as shown in the following Table 4.1:

TABLE 4.1 | ||

LX[4:3] + R[4:3] | C_{4} |
STAGE 4 rotation |

00 | 0001 | 0 positions |

01 | 0010 | 8 positions |

10 | 0100 | 16 positions |

11 | 1000 | 24 positions |

In addition to the rotation from Table 4.1, however, recall that the STAGE 4 CONTROLLER also responds to the CARRY that may be generated from the STAGE 3 CONTROLLER. If no such CARRY is generated for a given set of values of the left and right shift arguments, then the STAGE 4 rotation is as shown above in Table 4.1. However, if a CARRY is generated by the STAGE 3 CONTROLLER, then that carry is used to change the decoded value of C_{4 }so as to rotate its value one bit to the left in response to the CARRY, thereby accomplishing a same result that would be obtained if the CARRY were instead added by the STAGE 4 CONTROLLER into the sum of LX[4:3] and L[4:3]. Thus, the effect of the CARRY on the control signal C_{4 }is as shown in the following Table 4.2:

TABLE 4.2 | ||

LX[4:3] + R[4:3] | C_{4}, with no CARRY |
C_{4}, with CARRY |

00 | 0001 | 0010 |

01 | 0010 | 0100 |

10 | 0100 | 1000 |

11 | 1000 | 0001 |

To further appreciate the preceding discussion of the operation of the ROTATE STAGE 4, _{4}=0001. Further, from Table 4.1, the value of C_{4}=0001 gives rise to a right rotation of 0 bit positions. Accordingly,

From the preceding discussion of the operation of data rotate unit **200**, one skilled in the art may appreciate that it achieves a result situated in the same bit positions as achieved by the two independent shifts described earlier in connection with the example of _{1}, from _{1}, from

With **200** for the example of **200**, as is now explored in connection with

In the example of **200** the binary values are L=10000, LX=01111, and R=01000. These binary values and the responsive rotation of each ROTATE STAGE is now examined in connection with _{2}, from

From the above, the preferred embodiments have been shown to achieve, with respect to a data argument that is 32 bits by way of example, a result that includes equivalent bit positioning as is achieved by two independent and opposite-directions shifts, where the extent of the shifts is dictated by respective shift arguments. Rather than separate shifts of the entire data argument, the preferred embodiment instead uses multiple rotate stages, where the distance of rotation of each stage is controlled in response to only a portion of at least one of the two shift arguments. In other words, each of the ROTATE STAGEs 1 and 2 rotates the data argument only with respect to a portion, namely, bit [0], of the L and R shift arguments, respectively. Further, the ROTATE STAGE 3 rotates the data argument only with respect to a portion, namely, bits [2:1] of both the L and R shift arguments. Lastly, the ROTATE STAGE 4 rotates the data argument only with respect to a portion, namely, bits [4:3] of both the L and R shift arguments, as well as in response to a carry bit from STAGE 3. As further appreciated below by the preferred embodiment schematics for implementing the various stages, this breaking down of the shift arguments into portions and controlling respective stages with such portions permits for considerably improved throughput in that the completed shifting (or rotating) operations may be achieved faster than two independent shift operations, each of which shifts the data argument fully in response to the entire respective shift argument.

**210** of logic gates provides the logical equivalent of adding the respective least significant bits of these two portions, that is, it provides the outputs with respect to bits L[1] and R[1], while a set **220** of logic gates provides the logical equivalent of adding the respective most significant bits of these two portions, that is, it provides the outputs with respect to bits L[2] and R[2]. Note also with respect to the shift argument L that **210** and **220** are decoded by a decode set **230**, which thereby provides the 4-bit control signal, C_{3}, from the STAGE 3 CONTROLLER. Each of these logic sets is described below.

Looking to logic set **210**, its logic circuitry is now derived with respect to the functionality discussed above for the STAGE 3 CONTROLLER and further by focusing on the bits L[1] and R[1] processed by that set. First, noting that bits L[1] and R[1] are respective portions of what may be observed here as bit sets L[2:1] and R[2:1], then set **210** can be considered to be providing the equivalent of adding the least significant bit in each 2-bit portion L[2:1] and R[2:1]. Thus, each bit L[1] and R[1] corresponds to a decimal value of either 0 or 1, and the possible sums provided by adding L[1] and R[1] are either decimal 0, 1, or 2. Accordingly, three different states, corresponding to these three possible sums, may be realized by logic set **210**. Specifically, a sum of decimal 0 is obtained only when L[1] and R[1] both equal 0. Such conditions may be stated in logic form, as in the following Equation 2:

Sum 0 occurs for: LX[1] AND RX[1] Equation 2

Next, a sum of decimal 1 is obtained only when either L[1]=0 while R[1]=1, or when L[1]=1 while R[1]=0. Such conditions may be stated in logic form, as in the following Equation 3:

Sum 1 occurs for: (LX[1] AND R[1]) OR (L[1] AND RX[1]) Equation 3

Lastly, a sum of decimal 2 is obtained only when both L[1]=1 and R[1]=1. Such a condition may be stated in logic form, as in the following Equation 4:

Sum 2 occurs for: (L[1] AND R[1]) Equation 4

Having developed Equations 2 through 4, one skilled in the art will readily appreciate that they are implemented in logic set **210**, as further detailed below.

With respect to Equation 2, which recall corresponds to a sum equal to 0, it is implemented in the bottom of logic set **210**; specifically, an AND gate **210** _{A1}, having inverted inputs, is connected to receive bits L[1] and R[1], where one skilled in the art will recognize that the combination of inverted inputs with AND gate **210** _{A1 }provides a logical NOR operation and, thus, in implementation a single NOR gate may be used. These inputs are the same as those in Equation 2, although their complements are used because AND gate **210** _{A1 }has inverted inputs. Further in this regard, note that the use of inverted inputs is preferred, as is used in the other input gates of logic set **210** in order to provide sufficient drive power and remove the load with respect to each of the input bits. By way of convention, because AND gate **210** _{A1 }therefore represents the case when the least significant bits of bit sets L[2:1] and R[2:1] sum to 0, then the output of AND gate **210** _{A1 }is indicated as LS0, where the “LS” indicates the least significant bits and the “0” indicates the sum of 0.

With respect to Equation 3, which recall corresponds to a sum equal to 1, it is implemented in the middle of logic set **210**. Specifically, an AND gate **210** _{A2}, having inverted inputs, is connected to receive bits L[1] and RX[1]. These inputs are the same as the first two logic signals in Equation 4, where again complements are used because AND gate **210** _{A2 }has inverted inputs. Similarly, an AND gate **210** _{A3}, having inverted inputs, is connected to receive bits LX[1] and R[1], where these inputs are the same as the second two logic signals in Equation 4, where again complements are used because AND gate **210** _{A3 }has inverted inputs. Finally, to complete Equation 4, the output of AND gates **210** _{A2 }and **210** _{A3 }are connected to the input of an OR gate **201** _{01}. By way of convention, because OR gate **210** _{01 }therefore represents the case when the LSBs of bit sets L[2:1] and R[2:1] sum to 1, then the output of OR gate **210** _{O1 }is indicated as LS1. Lastly, note that the schematic of **210** _{A2 }and **210** _{A3 }are such that they abut the inputs of an OR gate **201** _{O1}; as known in the art, this depiction is intended to illustrate that these three devices may be implemented as a single so-called OR-AND-INVERT gate, which thereby provides a lesser delay as compared to separately constructed gates.

With respect to Equation 4, which recall corresponds to a sum equal to 2, it is implemented in the top of logic set **210**; specifically, an AND gate **210** _{A4}, having inverted inputs, is connected to receive bits LX[1] and RX[1]. These inputs are the same as those in Equation 4, although their complements are used because AND gate **210** _{A4 }has inverted inputs. Also, again the combination of AND gate **210** _{A4 }with inverted inputs may be implemented as a single NOR gate. Lastly, by way of convention, because AND gate **210** _{A4 }therefore represents the case when the LSBs of bit sets L[2:1] and R[2:1] sum to 2, then the output of AND gate **210** _{A4 }is indicated as LS2.

Looking to logic set **220**, its logic circuitry is now derived with respect to the functionality discussed above for the STAGE 3 CONTROLLER and further by focusing on the bits L[2] and R[2] processed by that set. First, note that bits L[2] and R[2] are also portions of the bit sets L[2:1] and R[2:1], where here logic set **220** can be considered to be providing the equivalent of adding the MSB in each 2-bit portion L[2:1] and R[2:1]. Further, because bits L[2] and R[2] are in the second bit position for bit sets L[2:1] and R[2:1], then when either bit L[2] or R[2] is set, it represents a decimal value of 2; thus, each bit L[2] or R[2] represents a decimal value of either 0 or 2, and the possible sums provided by adding L[2] and R[2] are either decimal 0, 2, or 4; however, when the sum reaches 4, it wraps around to represent a modulo 4 output (and as also handled through the separate generation of a carry bit, detailed later). Accordingly, there are only two possible different states, corresponding to these two possible sums, and they are realized by logic set **220**. Specifically, a sum of decimal 0 is obtained either when L[1] and R[1] both equal decimal 0, or when L[1] and R[1] both equal decimal 2 because this provides a value of 0 due to the wraparound. Such conditions may be stated in logic form, as in the following Equation 5:

Sum 0 occurs for: (LX AND RX) OR (L AND R) Equation 5

Next, a sum of decimal 2 is obtained only when either L[2]=0 while R[2]=1, or when L[2]=1 while R[2]=0. Such conditions may be stated in logic form, as in the following Equation 6:

Sum 2 occurs for:(LX[2] AND R[2]) OR (L[2] AND RX[2]) Equation 6

Having developed Equations 5 and 6, one skilled in the art will readily appreciate that they are implemented in logic set **220**, as further detailed below.

With respect to Equation 5, which recall corresponds to a sum equal to 0, it is implemented in the bottom of logic set **220**. Specifically, an AND gate **220** _{A1}, having inverted inputs, is connected to receive bits L[2] and R[2]. These inputs are the same as the first two logic signals in Equation 5, where again complements are used because AND gate **220** _{A1 }has inverted inputs. Similarly, an AND gate **220** _{A2}, having inverted inputs, is connected to receive bits LX[2] and RX[2], where these inputs are the same as the second two logic signals in Equation 5, where again complements are used because AND gate **220** _{A2 }has inverted inputs and the AND gate with inverted inputs can be implemented as a single NOR gate. Finally, to complete Equation 5, the output of AND gates **220** _{A1 }and **220** _{A2 }are connected to the input of an OR gate **220** _{O1}. By way of convention, because OR gate **220** _{O1}, therefore represents the case when the MSBs of bit sets L[2:1] and R[2:1] sum to 0, then the output of OR gate **220** _{O1}, is indicated as MS0.

With respect to Equation 6, which recall corresponds to a sum equal to 2, it is implemented in the top of logic set **220**. Specifically, an AND gate **220** _{A3}, having inverted inputs, is connected to receive bits L[2] and RX[2]. These inputs are the same as the first two logic signals in Equation 6, where again complements are used because AND gate **220** _{A3 }has inverted inputs. Similarly, an AND gate **220** _{A4}, having inverted inputs, is connected to receive bits LX[2] and R[2], where these inputs are the same as the second two logic signals in Equation 6, where again complements are used because AND gate **220** _{A4 }has inverted inputs. Finally, to complete Equation 6, the output of AND gates **220** _{A3 }and **220** _{A4 }are connected to the input of an OR gate **220** _{O2 }and, thus, the combination of the AND gates, with inverted inputs, followed by the OR gate, is preferably implemented as a single OR-AND-INVERT gate. By way of convention, because OR gate **220** _{O2 }therefore represents the case when the MSBs of bit sets L[2:1] and R[2:1] sum to 2, then the output of OR gate **220** _{O2 }is indicated as MS2.

Turning now to decode set **230**, it provides the 4-bit control signal, C_{3}, in response to the outputs from logic sets **210** and **220**; further in this regard, the signal outputs and inputs are labeled, but to simplify the illustration, the actual connections from sets **210** and **220** to decode set **230** are eliminated in **210**, its logic circuitry is now derived with respect to the functionality discussed above with respect to the control signal C_{3}. Specifically, recall that C_{3 }provides a 4-bit control signal, where only one of those 4 bits is enabled at a time, and the specific enabled bit corresponds to an indication for the ROTATE STAGE 3 to shift its input either 0, 2, 4, or 6 bits, and where the extent of the shift depends on the sum of L[2:1] and R[2:1]. Specifically, because the STAGE 3 CONTROLLER is a modulo 4 system, then the sum of its inputs are either decimal 0, 1, 2 or 3. Thus, each of these sums corresponds to a respective shift control in the ROTATE STAGE 3 and as achieved through the control signal C_{3}, of either 0, 2, 4, or 6 bits. Accordingly, to derive the logic of decode set **230**, each combination of inputs providing a sum of 0 should be decoded to assert a line C_{3.0 }of control signal C_{3}, thereby indicating a rotation of zero bit positions by the ROTATE STAGE 3, while each combination of inputs providing a sum of 1 should be decoded to assert a line C_{3.1 }of control signal C_{3}, thereby indicating a right rotation of two bit positions by the ROTATE STAGE 3, while each combination of inputs providing a sum of 2 should be decoded to assert a line C_{3.2 }of control signal C_{3}, thereby indicating a right rotation of four bit positions by the ROTATE STAGE 3, and lastly each combination of inputs providing a sum of 3 should be decoded to assert a line C_{3.3 }of control signal C_{3}, thereby indicating a right rotation of six bit positions by the ROTATE STAGE 3. Indeed, these various obtainable sums and the corresponding desired right rotations also may be characterized by the state of the least significant bit, LSn, and the most significant bit, MSn, from logic sets **210** and **200**. Particularly, the following Table 4.3 designates each possible state of these two bits, as well as the sum provided by those states and the desired value for the 4-bit control signal, C_{3}, corresponding to each sum:

TABLE 4.3 | |||||

LSn (decimal) | MSn (decimal) | Sum (decimal) | C_{3} | ||

0 | 0 | 0 | 0001 | ||

1 | 0 | 1 | 0010 | ||

2 | 0 | 2 | 0100 | ||

0 | 2 | 2 | 0100 | ||

1 | 2 | 3 | 1000 | ||

2 | 2 | 0 (modulo 4) | 0001 | ||

Each of these combinations is further derived below.

Looking toward the bottom of decode set **230**, it includes the logic for each combination of inputs providing a sum of 0, and which is thereby decoded to assert line C_{3.1 }of control signal C_{3 }to indicate a shift of zero positions by the ROTATE STAGE 3. Specifically, returning to set **220**, recall its output MS0 is asserted when its inputs total 0, and returning to set **210**, recall its output LS0 is asserted when its inputs total 0. Thus, when both of these outputs are asserted, then the sum of the least and most significant bits of L[2:1] and R[2:1] also equals zero and, hence, it is desirable to indicate a right rotation of zero positions, that is, to assert output C_{30.0}. Accordingly, such conditions may be stated in logic form, as in the following Equation 7a:

Sum 0 occurs for: LS0 AND MS0 Equation 7a

However, recalling that the STAGE 3 CONTROLLER is in part a modulo 4 system, then in addition to the condition of Equation 7a, when each of sets **210** and **220** provides an output of 2 (i.e., when LS2 and MS2 are asserted), then the sum corresponding to those two outputs equals 4, thereby wrapping back to a total of 0. Accordingly, Equation 7a may be re-written to add these additional conditions, as shown in the following Equation 7b:

Sum 0 occurs for: (LS0 AND MS0) OR (LS2 AND MS2) Equation 7b

Looking then to the devices that provide the output C_{3.0}, one skilled in the art will appreciate that they implement the logic of Equation 7b. Specifically, as to the first two logic states in Equation 7b, they are connected as inputs to an AND gate **230** _{A1}, and as to the second two logic states in Equation 7b, they are connected as inputs to an AND gate **230** _{A2}. The output of each of AND gates **230** _{A1 }and **230** _{A2 }is connected as an input to a NOR gate **230** _{N1}. Note, however, that Equation 7b calls for an OR function rather than a NOR as provided by NOR gate **230** _{N1}. Further in this regard, however, the output of NOR gate **230** _{N1 }is connected as input to an inverter **230** _{IN1}, which thereby in combination with NOR gate **230** _{N1 }provides an equivalent logical OR function. This combination of the AND gates and the NOR gate is preferably implemented as a single AND-OR-INVERT gate. As an additional benefit, however, the inclusion of inverter **230** _{IN1 }provides signal buffering for signal C_{3.0 }prior to its provision to the ROTATE STAGE 3. Similar buffering is provided for the remaining signals in control signal C_{3}, as further appreciated below.

Continuing with decode set **230**, it also includes the logic for each combination of inputs providing a sum of 1, and which are thereby decoded to assert line C_{3.1 }to indicate a right rotation of two bit positions by the ROTATE STAGE 3. Returning to sets **210** and **200** and as shown in Table 4.3, the only combination of an output from each that will total 1 occurs when both LS1 is asserted and MS0 is asserted, that is, at that time, the 2 LSBs of the sum of L[2:11] and R[2:1] equals 1 and, hence, it is desirable to indicate a right rotation of two bit positions by asserting output C_{3.1}. Accordingly, such conditions may be stated in logic form, as in the following Equation 8:

Sum 1 occurs for: LS1 AND MS0 Equation 8

Looking then to the devices that provide the output C_{3.1}, one skilled in the art will appreciate that they implement the logic of Equation 8. Specifically, the two logic states in Equation 8 are connected as inputs to a NAND gate **230** _{NA1}. Note, however, that Equation 8 calls for an AND function rather than a NAND as provided by NAND gate **230** _{NA1}. Further in this regard, however, the output of NAND gate **230** _{NA1 }is connected as input to an inverter **230** _{IN2}, which thereby in combination with NAND gate **230** _{NA1 }provides an equivalent logical AND function. Also, as introduced above, the inclusion of inverter **230** _{IN2 }provides signal buffering for signal C_{3.1 }prior to its provision to the ROTATE STAGE 3.

Continuing with decode set **230**, it also includes the logic for each combination of inputs providing a sum of 2, and which are thereby decoded to assert line C_{3.2 }to indicate a rotation of four bit positions by the ROTATE STAGE 3. Returning to sets **210** and **220** and as shown in Table 4.3, there are two combinations of an output from each that will sum to 2. Specifically, as a first combination, when MS0 is asserted, corresponding to an MSB total of 0, and when at the same time LS2 is asserted, corresponding to an LSB total of 2, then the sum of the two is decimal 2. Further, as a second combination, when MS2 is asserted, corresponding to an MSB total of decimal 2, and when at the same time LS0 is asserted, corresponding to an LSB total of 0, then the sum of the two is also decimal 2. In either case, it is desirable to indicate a right rotation of four bit positions by asserting output C_{3.2}. Accordingly, such conditions may be stated in logic form, as in the following Equation 9:

Sum 2 occurs for: (MS0 AND LS2) OR (MS2 AND LS0) Equation 9

Looking then to the devices that provide the output C_{3.2}, one skilled in the art will appreciate that they implement the logic of Equation 9. Specifically, the first two logic states in Equation 9 are connected as inputs to an AND gate **230** _{A3}, and the second two logic states in Equation 9 are connected as inputs to an AND gate **230** _{A4}. Further, the output of each of AND gates **230** _{A3 }and **230** _{A4 }is connected as an input to a NOR gate **230** _{N2}, which in combination with providing its output to an inverter **230** _{IN3}, provides the logical equivalent of the OR function from Equation 9, as well as a buffering aspect as discussed above. Additionally, this combination of the AND gates and the NOR gate is preferably implemented as a single AND-OR-INVERT gate.

Completing decode set **230**, it also includes the logic for each combination of inputs providing a sum of decimal 3, and which is thereby decoded to assert line C_{3.3 }to indicate a rotation of six bit positions by the ROTATE STAGE 3. Returning to sets **210** and **220** and as shown in Table 4.3, the only combination of an output from each that will total decimal 3 occurs when LS1 is asserted and MS2 is asserted, that is, at the time the 2 LSBs of the sum of L[2:1] and R[2:1] equals 3 and, hence, it is desirable to indicate a right rotation of six bit positions by asserting output C_{3.3}. Accordingly, such conditions may be stated in logic form, as in the following Equation 10:

Sum 3 occurs for: LS1 AND MS2 Equation 10

Looking then to the devices that provide the output C_{3.3 }and thereby implement the logic of Equation 10, the two logic states in Equation 10 are connected as inputs to a NAND gate **230** _{NA2}. Further, the output of NAND gate **230** _{NA2 }is connected to an inverter **230** _{IN4}, which thus in combination with NAND gate **230** _{NA2 }provides the logical equivalent of the AND function from Equation 10, as well as a buffering aspect as discussed above.

**240**, which is shown in greater detail in **240**, its logic circuitry is now derived with respect to the functionality discussed above for the STAGE 3 CONTROLLER and further by focusing on the bits L[2:1] and R[2:1] processed by that controller. Generally, the STAGE 3 CONTROLLER functions as a 2-bit adder (i.e., operable to add one 2-bit quantity with another 2-bit quantity), where as discussed above with respect to

TABLE 5 | ||||

Case | Addends, result | Carry status | ||

1 | 0 + 0 = 0 | None | ||

2 | 0 + 1 = 1 | None | ||

3 | 1 + 0 = 1 | None | ||

4 | 1 + 1 = 0 | Carry generate | ||

Table 5 illustrates that case 4, that is, when both 1-bit addends equal 1, is referred to as a carry generate because those addends alone give rise to a carry.

In addition to the notion of a carry generate as provided by Table 5, note also that cases 2 and 3 (i.e., addends of 0,1 or 1,0) are referred to in the art as a carry propagate, because if there is a carry generated in immediately lesser significant bits, then that same carry is propagated onward by cases 2 and 3. To further elaborate, for example, consider the addition in the following Equation 11, with bit positions [0] and [1]:

In Equation 12, case 3, a carry propagate, is shown in the MSBs at position [1]. However, the next lesser significant bits in position [0], namely the values of 0 and 0, do not generate a carry. Therefore, there is no carry to propagate onward by the case 3 bits. In contrast, consider now the following Equation 12:

In Equation 11, once again, a carry propagate is shown in the bit position [1]. However, here the next lesser significant bits, in position [0], namely the values of 1 and 1, do generate a carry. That carry is taken to, and shown vertically over, the bits **1** and **0** in position [1]; due to their value, however, those bits effectively propagate that carry upward to position [2]. Therefore, it can be seen how the propagate case 3 from Table 5 passes the carry bit to the next most significant bit position, and the same holds true for case 2 from Table 5.

Given the preceding, a logic equation now may be derived to indicate when a carry bit should be generated for an adder that adds two 2-bit quantities. First, Table 5 demonstrates that a carry is always generated when case 4 is provided. Second, Equation 12 demonstrates that a carry is also provided when case 2 or 3 is provided, and when at the same time a carry is generated in the immediately-lesser significant bit position. Thus, these conditions may be stated logically as in the following Equation 13:

CARRY=CASE 1+[(CASE 2 OR CASE3) AND (CASE 4 at preceding bit position)] Equation 13

Next, recalling that the STAGE 3 CONTROLLER is providing the equivalent to adding L[2:1] and R[2:1], and that carry generator **240** is therefore providing a carry for those bits, then these values may be substituted into Equation 13 to provide the following Equation 14:

*CARRY*=(*L[*2] AND *R[*2]) OR [(*L[*2] OR *R[*2]) AND (*L[*1] AND *R[*1])] Equation 14

Looking now to the schematic of **240** _{A1}; however, rather than connecting the output of AND gate **240** _{A1}, to an OR gate as shown in Equation 14, an equivalent function is realized by connecting the output of AND gate **240** _{A1}, as an input to a NOR gate **240** _{NO1}, where the DISABLE signal is also connected as an input to NOR gate **240** _{NO1}, and the output of NOR gate **240** _{NO1 }is connected as an input to a NAND gate **240** _{NA2}. Thus, when the inputs L[2] and R[2] as connected to AND gate **240** _{A1 }are true, then CARRY is asserted high, as will also occur when DISABLE is asserted. In an alternative manner, and where this alterative is the functional implementation of the first-listed OR function in Equation 14, the CARRY signal can be asserted high in response to an OR gate **240** _{OR1 }and a NAND gate **240** _{NA1}, where these two gates also are preferably implemented as a single complex gate. Specifically, L[2] and R[2] are connected as input to OR gate **240** _{OR1}, thereby implementing the elements (L[2] OR R[2]) of Equation 14, and the output of OR gate **240** _{OR1 }is connected as an input to NAND gate **240** _{NA1 }along with inputs R[1] and L[1]. Due to the NAND to NAND connection of gates **240** _{NA1 }and **240** _{NA2}, then when all of the inputs to NAND gate **240** _{NA1 }are true, then CARRY is asserted.

In Equation 15, the bits in position [0] generate a carry; however, the bits in position [1] are a carry kill. Thus, the example illustrates that the carry from position [0] does not pass beyond position [1], that is, due to the bit values in position [1], they are said to kill the carry. Accordingly, one condition in which case CARRYX should be asserted is when a kill occurs in the most significant bit position of the 2-bit addends, and thus, this condition may be stated according to the following Equation 16a:

*CARRYX*=(*LX[*2] AND *RX[*2]) Equation 16a

Further, however, one skilled in the art can confirm that a carry also will not be generated whenever at least one addend of both positions of the two bit positions is a zero. Thus, this condition may be stated as in the following Equation 16b:

*CARRYX*=[(*LX[*2] OR *RX[*2]) AND (*LX[*1] OR *RX[*1])] Equation 16b

Thus, by combining the conditions of Equations 16a and 16b, the following logic Equation 16c is formed to depict the instance when no carry occurs and, hence, when CARRYX should be asserted.

*CARRYX*=(*LX[*2] AND *RX[*2]) OR [(*LX[*2] OR *RX[*2]) AND (*LX[*1] OR *RX[*1])] Equation 16c

Returning now to the schematic of **240** _{A2}; however, rather than connecting the output of AND gate **240** _{A2 }to an OR gate as shown in Equation 16c, an equivalent function is realized by connecting the output of AND gate **240** _{A2 }as an input to a NOR gate **240** _{NO2}, where the DISABLE signal is also connected as an input to NOR gate **240** _{NO2}, and the output of NOR gate **240** _{NO2 }is connected as an input to a NAND gate **240** _{ONA3}. Thus, when the inputs LX[2] and RX[2] as connected to AND gate **240** _{A2 }are true, then CARRYX is asserted high, as will also occur when DISABLE is asserted. In an alternative manner, the CARRYX signal can be asserted high in response to high outputs from both an OR gate **240** _{OR2 }and an OR gate **240** _{OR3}. Specifically, LX[2] and RX[2] are connected as inputs to OR gate **240** _{OR2}, thereby implementing the elements (LX[2] OR RX[2]) of Equation 16c, and the output of OR gate **240** _{OR2 }is connected as an input to NAND gate **240** _{NA4}. Additionally, LX[1] and RX[1] are connected as inputs to OR gate **240** _{OR3}, thereby implementing the elements (LX[1] OR RX[1]) of Equation 16c, and the output of OR gate **240** _{OR3 }is connected as an input to NAND gate **240** _{NA4}. Due to the NAND to NAND connection of gates **240** _{NA4 }and **240** _{NA3}, then when all of the inputs to NAND gate **240** _{NA4 }are true, then CARRYX is asserted. Once more, the preceding gates may be consolidated and implemented as single complex gates.

**230** _{IN1}, **230** _{IN2}, **230** _{IN3}, and **230** _{IN4}), and to simplify **210**′-**230**′. However, in the preferred embodiment, the outputs of those blocks do not provide the ultimate control signal, C_{4}, for the ROTATE STAGE 4 because there also must be an accounting for any carry generated by the STAGE 3 CONTROLLER. Thus, the outputs of block **210**′-**230**′ are only preliminary signals and, hence, each is designated with an apostrophe to form a total signal C_{4}′ and to thereby indicate that it may be further changed in response to the CARRY and CARRYX signals, which also are shown as inputs to the STAGE 4 CONTROLLER. The remaining logic shown in

Recalling the relationship of the ROTATE STAGE 3 and the ROTATE STAGE 4 and their respective controllers, it has been shown that a carry in STAGE 3 should cause an increment in the sum of STAGE 4. However, as mentioned earlier, rather than having the CARRY signal from STAGE 3 change the sum of STAGE 4, instead in the preferred embodiment and to increase speed the carry is instead provided to change the decoded value of C_{4 }that results solely from the sum of LX[4:3] and R[4:3]. Thus, in _{4 }is originally provided in a decoded form in the signals C_{4.0}′, C_{4.1}′, C_{4.2}′, and C_{4.3}′, which without any other change would indicate to the ROTATE STAGE 4 a shift of 0, 8, 16, or 24 shift positions, respectively. Thus, in any one cycle, only one of those four signals C_{4.x}′ is asserted. Additionally, however, if CARRY is asserted, then these control signals are rotated once to the right so that the next number of rotations is effected as opposed to that indicated by the one asserted signal in C_{4.0}′, C_{4.1}′, C_{4.2}′, and C_{4.3}′. For example, if C_{4.0}′ is asserted corresponding to a rotation of 0 bit positions, and if CARRY is also asserted, then the assertion of C_{4.0}′ is effectively rotated right once to indicate a rotation of 8 bit positions. Similarly, if C_{4.0}′ is asserted corresponding to a rotation of 8 bit positions, and if CARRY is also asserted, then the assertion of C_{4.1}′ is effectively rotated right once to indicate a rotation of 16 bit positions. Also similarly, if C_{4.2}′ is asserted corresponding to a rotation of 16 bit positions, but if CARRY is also asserted, then the assertion of C_{4.2}′ is effectively rotated right once to indicate a shift of 24 bit positions. Lastly, therefore, if C_{4.3}′ is asserted corresponding to a rotation of 24 bit positions, but if CARRY is also asserted, then the assertion of C_{4.3}′ is used to indicate a rotation of 0 bit positions. Each of these rotations of the control signal is achieved by relocating the asserted signal C_{4.x}′ to the next highest conductor in the ultimate output signal C_{4}, that is, C_{4.x}′ is adjusted to be asserted as C_{4.x+1 }(i.e., without the apostrophe), with the additional wraparound such that C_{4.3}′ is adjusted to wraparound and thereby assert C_{4.0 }if CARRY is asserted.

To implement the above-described shift between C_{4}′ and C_{4}, the logic needed may be stated according to the following Equation 17, with the exception of the case when the bits in C_{4}′ wrap around due to the assertion of CARRY such that an asserted MSB in C_{4}′ wraps around to become an asserted LSB in C_{4}:

*C* _{4.x}=(*C* _{4.x}′ AND *CARRYX*)+(*C* _{4.x−1}′ AND *CARRY*) Equation 17

Equation 17a, therefore, demonstrates that when CARRY is not asserted (and hence CARRYX is asserted), then the output C_{4.x }is the same as the corresponding output C_{4.x}′ output from blocks **210**′-**230**′. Alternatively, when CARRY is asserted, then each output C_{4.x }is the next lesser significant bit C_{4.x−1}′ output from blocks **210**′-**230**′. Thus, the assertion of CARRY shifts the decoded value as desired and as discussed above. In addition, however, in the preferred embodiment, the signals for C_{4 }are negative logic (i.e., active low). Accordingly, to achieve negative logic signals the CARRY and CARRYX signals from Equation 17 are inverted. Each active low signal is also inverted locally inside the multiplexer of STAGE 4 that receives the C_{4.x }signal to generate the complementary signals for the transfer gates, so either polarity of select signal is acceptable.

One skilled in the art will readily appreciate that Equation 17 is implemented for each of the four signals to be used for C_{4}, also taking into account the wraparound case of C_{4 }as discussed above. For example, looking to the gates at the bottom of _{4.1}. From Equation 17, therefore, the inputs to a first AND gate would be C_{4.1}′ AND CARRYX. However, in the preferred embodiment, the AND gate provided is AND gate **250** _{AN1}, which has inverted inputs; thus, its inputs are CX_{4.1}′ AND CARRY. Similarly, the inputs to a second AND gate would be C_{4.4}′ AND CARRY. However, in the preferred embodiment, the AND gate provided is AND gate **250** _{AN2}, which also has inverted inputs; thus, its inputs are CX_{4.4}′ AND CARRYX. The output of each of AND gates **250** _{AN1 }and **250** _{AN2 }is connected as an input to an OR gate **250** _{OR1}, which therefore provides the C_{40 }signal; thereafter, that signal is inverted by an inverter **250** _{IN1}, providing an inverted output signal CX_{4.0}. Given these connections, one skilled in the art will appreciate the comparable devices and connections to provide each of CX_{4.1}, CX_{4.2}, and CX_{4.3}.

**100**′ according to an alternative preferred embodiment to processor **100** of **100**′ also may be a microprocessor or DSP or still other computing integrated circuit. Processor **100**′, like processor **100**, may include numerous aspects known in the art and that are neither shown nor described in order to simplify the remaining discussion. Processor **100**′ also includes an alternative data rotate unit **200**′, which shares many aspects with data rotate unit **200** of processor **100**. Thus, these common aspects are not detailed again as the reader is assumed to be familiar with the earlier discussion. Instead, the following concentrates on the differences between data rotate unit **200**′ and the above-described data rotate unit **200**. By way of introduction to these differences, in the preferred embodiment data rotate unit **200**′ is constructed using dynamic logic circuitry, as further detailed below in various schematics, where recall that data rotate unit **200** may be more readily implemented using static logic. As known in the dynamic logic circuitry art, dynamic logic operates in two phases, a precharge phase during which various precharge nodes are precharged to a first voltage, and an evaluate phase during which the data is read based on the voltage at the precharge node, where the precharge voltage therefore may be read as a first logic state if undisturbed or where that precharge voltage may be first discharged and then read as a second logic that that is complimentary to the first logic state; also in dynamic logic circuitry, typically dual rail signals are provided. Looking now to the overall block structure of data rotate unit **200**′ and its ROTATE STAGE 1′, its overall functionally is comparable to the ROTATE STAGE 1 of unit **200**, although it is constructed and operates using dynamic logic. However, the ROTATE STAGE 2′ of unit **200**′ combines the functionality of the ROTATE STAGEs 2 and 3 of unit **200** and, thus, as shown in _{3 }as well as the least significant bit of the L shift argument value, L[0] (and its complement, not separately shown). Note that this combination within ROTATE STAGE 2′ is not by way of limitation with respect to the inventive scope, but rather illustrates in block form how the preferred embodiment schematic is implemented to provided both functions as discussed below; accordingly, ROTATE STAGE 2′ also may be considered to comprise two stages, one directed to the sum of LX[2:1]+R[2:1] and the other directed to R[0]. Lastly, the overall functionality of the ROTATE STAGE 3′ is comparable to the ROTATE STAGE 4 of unit **200**, although it is constructed and operates using dynamic logic.

_{1IN }through D_{0IN }and outputting a 32-bit output having bits D_{31OUT }through D_{0OUT}. To simplify the discussion, note that the **300** _{0 }through **300** _{31}, each associated with a respective and different one of the outputs. Further, to simplify the illustration, only four of these sub-circuits are shown, where one skilled in the art will appreciate the comparable connections to the remaining circuits. The following discussion concentrates on only the illustrated circuits. Looking to sub-circuit **300** _{0}, it receives the data bit input D_{0IN }to an inverter **300** _{INV0}, which has its output connected to the source of an n-channel transistor **300** _{NCH0}. The gate of n-channel transistor **300** _{NCH0 }is connected to LX[0] and the source of that same transistor is connected to a node **300** _{N0}. Node **300** _{N0 }is also connected to the drain of a p-channel transistor **300** _{PCH0}, which has its source connected to a reference voltage, V_{DD}, and its gate connected to receive a clock signal CLK. Node **300** _{N0 }is also connected to the drain of an n-channel transistor **300** _{NCH1}, which has its gate connected to L[0] and its source connected to the output of an inverter **300** _{INV2 }from sub-circuit **300** _{1}. Completing sub-circuit **300** _{0}, node **300** _{N0 }is connected to an input of an inverter **300** _{INV1}, and the output of inverter **300** _{INV1 }provides the output signal D_{0OUT}. Given the above connections, one skilled in the art will readily appreciate that each remaining sub-circuit **300** _{x }in _{xIN }and to provide a corresponding output D_{xOUT}. Lastly, note that

The operation of the ROTATE STAGE 1′, as implemented using the sub-circuits of _{31IN }through D_{0IN}, is passed directly to outputs D_{31OUT }through D_{0OUT }if LX[0] is true, whereas if LX[0] is false (i.e., L[0] is true), then the input is shifted once to the right such that each input D_{xIN }is connected to an output D_{x-1OUT}, and the LSB D_{0IN }thereby wraps around to be output by D_{31OUT}. This operation is now examined in greater detail, and by looking to sub-circuit **300** _{0 }by way of example. First, during a precharge phase of operation, CLK goes low. In response, p-channel transistor **300** _{PCH0 }is enabled, and node **300** _{N0 }is precharged. Thus, also during this precharge phase, the comparable p-channel transistors **300** _{PCHx }in the remaining sub-circuits of **300** _{Nx}. Second, during an evaluate phase of operation, CLK goes high. At this time, the voltage at each node **300** _{Nx }either remains in its precharged phase or it is discharged, where the result depends on the value of LX[0] as well as the input data. Specifically, looking to sub-circuit **300** _{0}, if LX[0] is true, then node **300** _{N0 }is connected through n-channel transistor **300** _{NCH0 }to the output of inverter **300** _{INV0}. Accordingly, if the output of inverter **300** _{INV0 }is low (i.e., because its input, D_{0IN }is high), then the precharged voltage at node **300** _{N0 }is discharged, and in response a rising voltage is created at the output of inverter **300** _{INV1 }to the output D_{0OUT}. Conversely, if the output of inverter **300** _{INV0 }is high (i.e., because its input, D_{0IN }is low), then the precharged high voltage at node **300** _{N0 }is undisturbed and, thus, inverter **300** _{INV1 }continues to output a low voltage to the output D_{0OUT}. The two alternatives just provided are for the case when LX[0] is true; however, if LX[0] is false, then its complement L[0] is true, thereby enabling n-channel transistor **300** _{NCH0 }(while n-channel transistor **300** _{NCH0 }remains disabled). In this case, the data to sub-circuit **300** _{0 }is not provided by input D_{0IN}, but instead it is provided by input D_{1IN}, and in inverted form due to inverter **300** _{INV2}. Thus, when LX[0] is false, there is a shift of data one bit position to the right. With the data as provided by input D_{1IN}, it will either discharge node **300** _{N0 }if that data is high and thus inverted to a low via inverter **300** _{INV2}, or it will leave the precharge voltage on node **300** _{N0 }undisturbed if that data is low and thus inverted to a high via inverter **300** _{INV2}.

_{3}, and another rotation possibility of one bit position occurs based on the value of L[0]. Thus, multiplexer M2′ right rotates its input from 1 to 8 bit positions. Each of these aspects is further detailed below. Lastly, in

Looking more specifically to the connections and devices of multiplexer M2′, it includes four data inputs d1, d3, d5, and d7, where the “d” indicates data and the number indicates a relative bit offset from the multiplexer. Looking first to input d1, it has an offset of one bit position, meaning this input receives a bit that is right-rotated one bit position relative to the output from ROTATE STAGE 1′, and it is then provided as an input d1to multiplexer M2′. This one bit position rotation, therefore, is a hardwire implementation of the “+1” aspect of Equation 1, that is, for all data bits output from ROTATE STAGE 1′ to ROTATE STAGE 2′, each of those bits is necessarily right-rotated one bit position. Continuing with input d3, it has an offset of three bit positions, meaning this input receives a bit that is right-rotated three bit positions relative to the output from ROTATE STAGE 1′. One of these three bit positions accommodates the hardwiring of the “+1” aspect of Equation 1, while the remaining two bit positions represent a potential right-rotation due to the state of C_{3 }and, more particularly, due to the state of C_{3.2}. In a comparable manner but for greater potential offsets, input d5 has an offset of five bit positions relative to the output from ROTATE STAGE 1′ and input d7 has an offset of seven bit positions relative to the output from ROTATE STAGE 1′. Lastly, dual rail inputs dXn are provided for each respective input dn.

Each data input dn is connected to a first source/drain of a respective n-channel transistor **310** _{dn}, and each data input dXn is connected to a first source/drain of a respective n-channel transistor **310** _{dXn}. The gate of n-channel transistor **310** _{d}1is connected to receive the control signal C_{3.0}, which is intended to serve the same function as provided from the static embodiment **310** _{d}3is connected to receive the control signal C_{3.1}, which corresponds to a right rotation of 2 additional bit positions. The gate of n-channel transistor **310** _{d}5is connected to receive the control signal C_{3.2}, which corresponds to a right rotation of 4 additional bit positions. The gate of n-channel transistor **310** _{d}7is connected to receive the control signal C_{3.3}, which corresponds to a right rotation of 6 additional bit positions. The second source/drain of each n-channel transistor **310** _{dn }is connected to a node **312**. Symmetric connections are made for each dual-rail input dXn to a first source/drain of a corresponding n-channel transistor **310** _{dXn}, and the second source/drain of each n-channel transistor **310** _{dXn }is connected to a node **312**X.

Node **312** is connected to a first source/drain of an n-channel transistor **314** that has its gate connected to bit RX[0] of the R shift argument value, and the second source drain of n-channel transistor **314** is connected to a node **316**. Node **316** is also connected to a first source/drain of an n-channel transistor **318** that has its gate connected to bit R[0] of the R shift argument value and its second source/drain connected to a data bit input d_{m+1}. Data bit input d_{m+1 }is intended to depict an output from the multiplexer just to the left of multiplexer M2′ (not expressly shown), where that specific output is from the first set of n-channel transistors that are comparable to n-channel transistors **310** _{d}1through **310** _{d}7in _{m+1 }can be selected to achieve an additional right rotation of one bit position, so to achieve that one bit rotation the data is provided from the multiplexer just to the left of multiplexer M2′; however, input d_{m+1 }is not from the final output of that to-the-left multiplexer, but instead it is taken from the node that is comparable to node **312** in **312** provides an output out_{m }(and node **312**X provides a dual rail output out_{mX}). This output is therefore available to the multiplexer that is immediately to the right of multiplexer M2′ to serve a comparable value to that received by the illustrated input d_{m+1}. As a result of this connection, the one-bit right rotation achieved via n-channel transistor **318** is in addition to either a 0, 2, 4, or 6 bit right rotation achieved via the first group of n-channel transistors in the multiplexer immediately to the left of multiplexer M2′. Node **316** is also connected to the drain of a p-channel transistor **320** which has its gate connected to the CLK signal and its source connected to V_{DD}. Further, node **316** is connected to the input of an inverter **322** that has its output connected to the input of an inverter **324**. The output of inverter **324** provides one of the dual rail outputs of multiplexer M2′.

The operation of multiplexer M2′ is now described, and it might well be readily understood given the various functional descriptions provided above with respect to the connections of the Figure as well as the understood operation of the ROTATE STAGE 2′. During the precharge phase, CLK is low and p-channel transistors **320** and **320**X are enabled, thereby precharging nodes **316** and **316**X, respectively. During the evaluate phase, CLK transitions high, and in response to the various select signals and the data, one of these two precharged nodes **316** and **316**X is discharged. First, recall that one of the four signals C_{3.x }of C_{3 }is asserted, and that asserted signal enables the n-channel transistor **310** _{dx+1 }to which it is connected; the data dn connected to the enabled transistor is thereby connected to node **312** and is also available at the output out_{m }to the multiplexer located at one bit position to the right of multiplexer M2′ (while the complementary data dXn is connected to node **312**X and which also provides out_{mx }to the multiplexer located at one bit position to the right of multiplexer M2′). Next, the state of R[0] enables only one of n-channel transistors **314** or **318**, that is, if RX[0] is true, then n-channel transistor **314** is enabled, whereas if RX[0] is false, then its complement R[0] enables n-channel transistor **318**. In the former case, the data at node **312** is connected to node **316**, and in the latter case the data, d_{m+1 }is connected to node **316**. In either event, if the connected data is low, it thereby discharges the precharge voltage at node **316**. Finally, the voltage at node **316**, whether remaining precharged or whether discharged, is inverted twice and provided as the output out_{d+1}, thereby presenting an active low signal.

Looking more specifically to the connections and devices of multiplexer M3′, it includes four data inputs d0, d8, dl6, and d24, where the “d” indicates data and the number indicates a relative bit offset from the multiplexer. More particularly, input d0 has no offset from multiplexer M3′, but input d8 is data that is offset by 8 bit positions so that it can be selected to achieve a right rotation of 8 bit positions. In other words, the input d8 is the same bit that would be an input d0 to another multiplexer in ROTATE STAGE 3′ that is 8 bit positions to the left of multiplexer M3′. Similarly, therefore, input d16 is data that is offset by 16 bit positions so that it can be selected to achieve a right rotation of 16 bit positions, and input d24 is data that is offset by 24 bit positions so that it can be selected to achieve a right rotation of 24 bit positions. Lastly, dual rail inputs dXn are provided for each respective input dn. In addition to the preceding inputs, both rails also include two additional inputs, d_{alt1 }and d_{alt2 }(with respective dual rail signals dX_{alt1 }and dX_{alt2}). These two additional inputs are included to demonstrate that multiplexer M3′ may perform with respect to inputs that are not provided by the preceding ROTATE STAGE 2′, but rather, are provided from some other circuit that still benefits from the functionality provided by multiplexer M3′.

Each data input dn is connected to a first source/drain of a respective n-channel transistor **410** _{dn}, and each data input dXn is connected to a first source/drain of a respective n-channel transistor **410** _{dXn}. The gate of n-channel transistor **410** _{d}0is connected to receive the control signal C_{4.0}, which is intended to serve the same function as provided from the static embodiment **410** _{d}8is connected to receive the control signal C_{4.1}, which corresponds to a right rotation of 8 bit positions. The gate of n-channel transistor **410** _{d}16is connected to receive the control signal C_{3.2}, which corresponds to a right rotation of 16 bit positions. The gate of n-channel transistor **310** _{d}24is connected to receive the control signal C_{3.3}, which corresponds to a right rotation of 24 bit positions. The second source drain of each n-channel transistor **410** _{dn }is connected to a node **412**. Symmetric connections are made for each dual-rail input dXn to a first source/drain of a corresponding n-channel transistor **410** _{dXn}, and the second source/drain of each n-channel transistor **410** _{dXn}, is connected to a node **412**X. Also to contrast two different approaches, recall there are two additional data inputs, d_{alt1 }and d_{alt2}. The input d_{alt1 }is a gate-driving data bit, that is, it is connected to the gate of an n-channel transistor **410** _{dalt1 }that has its source connected to ground and its drain connected to a first source/drain of an n-channel transistor **410** _{alt1}. The gate of n-channel transistor **410** _{alt1 }is connected to s_{alt1}, and the second source/drain of n-channel transistor **410** _{alt1 }is connected to node **412**. The input d_{alt2 }is a source-driving data bit, that is, it is connected to the source of an n-channel transistor **410** _{dalt2 }that has its gate connected to s_{alt2 }and its drain connected to node **412**.

The operation of multiplexer M3′ is now described, although it should be readily understood from the comparable operation of the front end portion of multiplexer M2′ described above as well as the understood operation of the ROTATE STAGE 3′. When the preceding ROTATE STAGE 2′ performs its evaluate phase, it passes active low data to multiplexer M3′ of ROTATE STAGE 3′. These inputs are multiplexed together along with the additional inputs d_{alt1 }and d_{alt2}, that is, from the entire set of inputs, one at a time may be selected, either by the assertion of one of the signals C_{4.x }or s_{altx}. Further in this regard, therefore, note that during a cycle wherein one of inputs d_{alt1 }and d_{alt2 }is to be selected, then none of the signals C_{4.x }is asserted. In any event, therefore, the one input selected during a clock cycle and of the various input signals is connected to node **412**, and its inverse is connected to node **412**X.

_{1}, L_{2}, L_{3}, and L_{4 }of transistors, where such a structure is further appreciated also with the operational description provided later.

Starting at the bottom of _{1}, the schematic includes two n-channel transistors, each having a source connected to a node **420**N, which is connected to ground. The gate of n-channel transistor **420** _{1 }is connected to RX[1] and its drain is connected to a node **422**N**1**. The gate of n-channel transistor **420** _{2 }is connected to R[1] and its drain is connected to a node **422**N**2**.

The next level upward from the bottom of _{2}, includes four n-channel transistors **422** _{1}, **422** _{2}, **422** _{3}, and **422** _{4}. The sources of n-channel transistors **422** _{1 }and **422** _{2 }are connected to node **422**N**1**, and the sources of n-channel transistors **422** _{3 }and **422** _{4 }are connected to node **422**N**2**. The gates of n-channel transistors **422** _{1 }and **422** _{3 }are connected to LX[1], and the gates of n-channel transistors **422** _{2 }and **422** _{4 }are connected to L[1]. The drain of n-channel transistor **422** _{1 }is connected to a node **424**N**1**, the drains of n-channel transistors **422** _{2 }and **422** _{3 }are connected to a node **424**N**2**, and the drain of n-channel transistor **422** _{3 }is connected to a node **424**N**3**.

The next level upward from the bottom of _{3}, includes six n-channel transistors **424** _{1}, **424** _{2}, **424** _{3}, **424** _{4}, **424** _{5}, and **424** _{6}. The sources of n-channel transistors **424** _{1}, and **424** _{2 }are connected to node **424**N**1**, the sources of n-channel transistors **424** _{3 }and **424** _{4 }are connected to node **424**N**2**, and the sources of n-channel transistors **424** _{5 }and **424** _{6 }are connected to node **424**N**3**. The gates of n-channel transistors **424** _{1}, **424** _{3}, and **424** _{5 }are connected to RX[2], and the gates of n-channel transistors **424** _{2}, **424** _{4}, and **424** _{6 }are connected to R[2]. The drain of n-channel transistors **424** _{1 }and **424** _{6 }are connected to a node **426**N**1**, the drains of n-channel transistors **424** _{2 }and **424** _{5 }are connected to a node **426**N**4**, the drain of n-channel transistor **424** _{3 }is connected to a node **426**N**2**, and the drain of n-channel transistor **424** _{4 }is connected to a node **426**N**3**.

The top level of _{4}, includes eight n-channel transistors **426** _{1}, **426** _{2}, **426** _{3}, **426** _{4}, **426** _{5}, **426** _{6}, **426** _{7}, and **426** _{8}. The gates of n-channel transistors **426** _{1}, **426** _{3}, **426** _{5}, and **426** _{7 }are connected to LX[2], and the gates of n-channel transistors **426** _{2}, **426** _{4}, **426** _{6}, and **426** _{8 }are connected to L[2]. The sources of n-channel transistors **426** _{1 }and **426** _{2 }are connected to node **426**N**1**, the sources of n-channel transistors **426** _{3 }and **426** _{4 }are connected to node **426**N**2**, the sources of n-channel transistors **426** _{5 }and **426** _{6 }are connected to node **426**N**3**, and the sources of n-channel transistors **426** _{7 }and **426** _{8 }are connected to node **426**N**4**. The drains of n-channel transistors **426** _{1}, and **426** _{8 }are connected to a node **428**N**1**, the drains of n-channel transistors **426** _{3 }and **426** _{6 }are connected to a node **428**N**2**, the drains of n-channel transistors **426** _{2 }and **426** _{7 }are connected to a node **428**N**3**, and the drains of n-channel transistors **426** _{4 }and **426** _{5 }are connected to a node **428**N**4**. Each of nodes **428**N**1**, **428**N**2**, **428**N**3** and **428**N**4** is connected to a drain of a respective p-channel transistor **430** _{1}, **430** _{2}, **430** _{3 }and **430** _{4}, and the gate of each of p-channel transistors **430** _{1}, **430** _{2}, **430** _{3 }and **430** _{4 }is connected to CLK while the sources of those p-channel transistors are connected to V_{DD}. Lastly, each of nodes **428**N**1**, **428**N**2**, **428**N**3** and **428**N**4** is connected to an input of a respective inverter **432** _{1}, **432** _{2}, **432** _{3 }and **432** _{4}, where the outputs of those inverters provide the respective signals C_{3.0}, C_{3.1}, C_{3.2}, and C_{3.3}.

The operation of the schematic of _{3.1 }is asserted, if the modulo 4 sum is 1 then output C_{3.2 }is asserted, if the modulo 4 sum is 2 then output C_{33 }is asserted, and if the sum is 3 then output C_{3.4 }is asserted. By way of further introduction, the following Table 6 illustrates the sum for each possible input of L[2:1] and R[2:1], given the decimal values of L[2:1] and R[2:1].

TABLE 6 | ||||

L[2:1] = 0 | L[2:1] = 1 | L[2:1] = 2 | L[2:1] = 3 | |

R[2:1] = 0 | 0 | 1 | 2 | 3 |

R[2:1] = 1 | 1 | 2 | 3 | 0 |

R[2:1] = 2 | 2 | 3 | 0 | 1 |

R[2:1] = 3 | 3 | 0 | 1 | 2 |

For further demonstrating the operation of the schematic of _{1 }of **420** _{1 }being enabled because R[1]=0 and thus RX[1]=1, at which point it couples node **420**N to node **422**N**1**. Alternatively in _{1}, then if the input bit R[1] equals 1, then the tree structure conducts toward the right path, which at that point indicates that the one input then considered, namely R[1], has provided a total as shown in the circle at the end of the path, that is, equal to 1. Looking back to _{1 }of **420** _{2 }being enabled because R[1]=1, at which point it couples node **420**N to node **422**N**2**. This same flow continues for each additional level, taking into account the additional one-bit input for that level, with level L_{2 }considering L[1], level L_{3 }considering R[2], and level L_{4 }considering L[2]. Various examples are provided below to further demonstrate this operation.

Consider the case of Table 6 where L[2:1]=R[2:1]=00 (i.e., both equal decimal zero), as with respect to _{1}, R[1]=0, and thus flow continues along the left branch to the level L_{1 }circle with a sum of 0. Next, at level L_{2}, L[1]=0, and thus flow continues along the left branch to the level L_{2 }circle with a sum of 0. Thus, at this point, the LSB of both L[2:1] and R[2:1] have been summed to the value of 0, as shown in the level L_{2 }circle with a sum of 0. Next, at level L_{3}, R[2]=0, and thus flow continues along the left branch to the level L_{3 }circle with a sum of 0. Finally, at level L_{4}, L[2]=0, and thus flow continues along the left branch to the level L_{4 }circle with a sum of 0. Looking back to **428**N**1**, **428**N**2**, **428**N**3**, and **428**N**4**; then in the evaluate phase, and in the same relative manner as was described with respect to **420** _{1}, **422** _{1}, **424** _{1}, and **426** _{1}. Thus, node **428**N**1** is discharged through this path, and that falling signal is inverted by inverter **432** _{1}, thereby providing an asserted high signal at C_{3.0}, while the other outputs C_{3.1}, C_{3.2}, and C_{3.3 }remain low (because the respective precharge nodes for those signals remain charged and inverted by the respective inverters **423** _{2}, **423** _{3}, and **432** _{4}).

Now consider the case of Table 6, where L[2:1]=00 (i.e., decimal 0) and R[2:1]=11 (i.e., decimal 3), as with respect to _{1}, R[1]=1, and thus flow continues along the right branch to the level L_{1 }circle with a sum of 1. Thus, at this point, the one bit that has been considered totals 1. Next, at level L_{2}, L[1]=0, and thus flow continues along the left branch to the level L_{2 }circle with a sum of 1. Thus, at this point, the LSB of both L[2:1] and R[2:1] have been summed to the value of 1, as shown in the level L_{2 }circle with a sum of 1. Next, at level L_{3}, R[2]=1, and thus flow continues along the right branch to the level L_{3 }circle with a sum of 3; note here that because this level is considering the second bit in the shift argument R, then that bit represents a value times 2, and since the bit equals one then it represents a decimal value of 2; accordingly, this decimal value of 2 added with the previous value of 1 provides a total of 3, as shown in the level L_{3 }circle. Finally, at level L_{4}, L[2]=0, and thus flow continues along the left branch to the level L_{4 }circle with a sum of 3. Looking to the corresponding operation in **428**N**1**, **428**N**2**, **428**N**3**, and **428**N**4** are precharged, then in the evaluate phase, and in the same relative manner as was described with respect to **420** _{2}, **422** _{3}, **424** _{4}, and **426** _{5}. Thus, node **428**N**4** is discharged through this path, and that falling signal is inverted by inverter **432** _{4}, thereby providing an asserted high signal at C_{3.3}, while the other outputs C_{3.0}, C_{3.1}, and C_{3.2 }remain low.

As a final example, consider the case of Table 6, where L[2:1]=01 (i.e., decimal 1) and R[2:1]=11 (i.e., decimal 3); in a decimal total, this sum equals 4, but recall the encoder is modulo 4 and, hence, the sum provided should be 0. Such a sum and its operation with respect to _{1}, R[1]=1, and thus flow continues along the right branch to the level L_{1 }circle with a sum of 1. Thus, at this point, the one bit that has been considered totals 1. Next, at level L_{2}, L[1]=1, and thus flow continues along the right branch to the level L_{2 }circle with a sum of 2. Thus, at this point, the LSB of both L[2:1] and R[2:1] have been summed to the value of 2, as shown in the level L_{2 }circle with a sum of 2. Next, at level L_{3}, R[2]=1, and thus flow continues along the right branch to the level L_{3 }circle with a sum of 0; note here that because this level has added 2 to the previous sum of 2, then the result of 4 is returned modulo 4 to a value of 0, as shown in the level L_{3 }circle. Finally, at level L_{4}, L[2]=0, and thus flow continues along the left branch to the level L_{4 }circle with a sum of 0. Looking to the corresponding operation in **428**N**1**, **428**N**2**, **428**N**3**, and **428**N**4** are precharged, then in the evaluate phase, and in the same relative manner as was described with respect to **420** _{2}, **422** _{4}, **424** _{6}, and **426** _{1}. Thus, node **428**N**1** is discharged through this path, and that falling signal is inverted by inverter **432** _{1}, thereby providing an asserted high signal at C_{3.0}, while the other outputs C_{3.1}, C_{3.2}, and C_{3.3 }remain low.

**15**, that is, to generate both the CARRY and CARRYX signals for use by the STAGE 3′ CONTROLLER. In general, **440** that has its source connected to ground, its gate connected to a control signal carry_enable, and its drain connected to a node **442**N. Node **422**N is also connected to a drain of a p-channel transistor **444** that has its gate connected to CLK and its source connected to V_{DD}.

Turning to the circuitry for generating the CARRY signal, it includes a p-channel transistor **446** having a source connected to V_{DD}, a gate connected to CLK, and a drain connected to a node **446**N, which is connected to the input of an inverter **448**. The output of inverter **448** provides the CARRY signal. Two n-channel transistors **448** _{1 }and **448** _{2 }have their source/drain conductive paths connected in series between nodes **446**N and **442**N, and their respective gates are connected to R[2] and L[2]. An additional p-channel transistor **450** _{1 }has its source connected to V_{DD}, its gate connected to CLK, and its drain connected to a node **450**N. Two n-channel transistor **450** _{2 }and **450** _{3 }have their source/drain conductive paths connected in parallel between nodes **450**N and **446**N, with the gate of n-channel transistor **450** _{2 }connected to L[2] and the gate of n-channel transistor **450** _{3 }connected to R[2]. Two n-channel transistors **450** _{4 }and **450** _{5 }have their source/drain conductive paths connected in series between nodes **450**N and **442**N, and their respective gates are connected to L[1] and R[1].

The operation of the circuitry for generating the CARRY signal in **448** _{1 }and **448** _{2 }provide the function of (L[2] AND R[2]) in that, after the precharge phase wherein node **446**N is precharged and assuming the carry_enable is asserted, then node **446**N is discharged during the evaluate phase if that function is met, thereby enabling those two n-channel transistors. Alternatively, n-channel transistors **450** _{2}, **450** _{3}, **450** _{4}, and **450** _{5 }provide the function of [(L[2] OR R[2]) AND (L[1] AND R[1])], where if either n-channel transistor **450** _{2 }or **450** _{3 }is enabled while at the same time both n-channel transistors **450** _{4 }and **450** _{5 }are enabled, then node **446**N is discharged. In any event, if node **446**N is discharged, then the output of inverter **448** rises, thereby asserting the CARRY signal.

Turning to the circuitry for generating the CARRYX signal, it includes a p-channel transistor **452** having a source connected to V_{DD}, a gate connected to CLK, and a drain connected to a node **452**N, which is connected to the input of an inverter **454**. The output of inverter **454** provides the CARRYX signal. Two n-channel transistors **456** _{1 }and **456** _{2 }have their source/drain conductive paths connected in series between nodes **452**N and **442**N, and their respective gates are connected to RX[2] and LX[2]. An additional p-channel transistor **458** _{1 }has its source connected to V_{DD}, its gate connected to CLK, and its drain connected to a node **458**N. Two n-channel transistor **458** _{2 }and **458** _{3 }have their source/drain conductive paths connected in parallel between nodes **458**N and **442**N, with the gate of n-channel transistor **458** _{2 }connected to LX[1] and the gate of n-channel transistor **458** _{3 }connected to RX[1]. Two n-channel transistor **458** _{4 }and **458** _{5 }have their source/drain conductive paths connected in parallel between nodes **452**N and **458**N, with the gate of n-channel transistor **458** _{4 }connected to LX[2] and the gate of n-channel transistor **458** _{5 }connected to RX[2].

The operation of the circuitry for generating the CARRYX signal in **456** _{1 }and **456** _{2 }provide the function of (LX[2] AND RX[2]) in that, after the precharge phase wherein node **452**N is precharged and assuming the carry_enable is asserted, then node **452**N is discharged during the evaluate phase if that function is met, thereby enabling those two n-channel transistors. Alternatively, n-channel transistors **458** _{2}, **458** _{3}, **458** _{4}, and **458** _{5 }provide the function of [(LX[2] OR RX[2]) AND (LX[1] OR RX[1])], where if either n-channel transistor **458** _{2 }or **458** _{3 }is enabled while at the same time either n-channel transistor **458** _{4 }or **458** _{5 }is enabled, then node **452**N is discharged. In any event, if node **452**N is discharged, then the output of inverter **454** rises, thereby asserting the CARRYX signal.

**432** _{1 }through **432** _{4}) are designated C_{4.0}″, C_{4.1}″, C_{4.2}″, and C_{4.3}″; thus, only one of those outputs is asserted at a time and in response to the modulo 4 sum of L[4:3]+R[4:3]. As shown in _{4.0}″, C_{4.1}″, C_{4.2}″, and C_{4.3}″ are provided as inputs to the remaining circuitry therein.

Continuing with the schematic of _{4.0}″, C_{4.1}″, C_{4.2}″, and C_{4.3}″ is connected to the gate of a respective n-channel transistor **470** _{0}, **470** _{1}, **470** _{2}, and **470** _{3}, where each of those n-channel transistors has its source connected to ground and its drain connected to a respective node **472**N_{0}, **472**N_{1}, **472**N_{2}, and **472**N_{3}. Node **472**N_{0 }is connected to the source of an n-channel transistor **474** _{0 }and the source of an n-channel transistor **474** _{1}, node **472**N_{1 }is connected to the source of an n-channel transistor **474** _{2 }and the source of an n-channel transistor **474** _{3}, node **472**N_{2 }is connected to the source of an n-channel transistor **474** _{4 }and the source of an n-channel transistor **474** _{5}, and node **472**N_{3 }is connected to the source of an n-channel transistor **474** _{6 }and the source of an n-channel transistor **474** _{7}. The gates of each of n-channel transistors **474** _{0}, **474** _{2}, **474** _{4}, and **474** _{6 }are connected to CARRYX, and the gates of each of n-channel transistors **474** _{1}, **474** _{3}, **474** _{5}, and **474** _{7 }are connected to CARRY. The drains of n-channel transistors **474** _{0 }and **474** _{7 }are connected to a node **476**N_{0}, the drains of n-channel transistors **474** _{1 }and **474** _{2 }are connected to a node **476**N_{1}, the drains of n-channel transistors **474** _{3 }and **474** _{4 }are connected to a node **476**N_{2}, and the drains of n-channel transistors **474** _{5 }and **474** _{6 }are connected to a node **476**N_{3}. Each node **476**N_{0}, **476**N_{1}, **476**N_{2}, and **476**N_{3 }is connected to the drain of a respective p-channel transistor **478** _{0}, **478** _{1}, **478** _{2}, and **478** _{3}, and the drain of each of those p-channel transistors is also connected to an input of a respective inverter **480** _{0}, **480** _{1}, **480** _{2}, and **480** _{3}. The gates of all of those p-channel transistors **478** _{0}, **478** _{1}, **478** _{2}, and **478** _{3 }are connected to CLK, and the sources of all of those p-channel transistors are connected to V_{DD}. The outputs of inverters **480** _{0}, **480** _{1}, **480** _{2}, and **480** _{3 }provides respective control signals C_{4.0}, C_{4.1}, C_{4.2 }and C_{4.3}, to thereby control the ROTATE STAGE 3′ as described above.

The operation of the schematic of **478** _{0}, **478** _{1}, **478** _{2}, and **478** _{3 }is enabled and thereby precharges its respective drain-connected node **476**N_{0}, **476**N_{1}, **476**N_{2}, and **476**N_{3}. Next, during the evaluate phase, only one of C_{4.0}″, C_{4.1}″, C_{4.2}″, and C_{4.3}″ is asserted in response to the addition of L[4:3]+R[4:3] as implemented in the multi-level structure (not explicitly shown). Consequently, the asserted one of C_{4.0}″, C_{4.1}″, C_{4.2}″, and C_{4.3}″ enables the one of n-channel transistors **470** _{0}, **470** _{1}, **470** _{2}, and **470** _{3 }to which it is connected. Further, however, either the CARRY or CARRYX signal will be asserted during the evaluate phase. If the CARRYX signal is asserted, then each node **476**N_{1 }is connected by the respective n-channel transistor **474** _{x }to a respective node **472**N_{x}. However, if the CARRY signal is asserted, then note that the illustrated structure rotates the 4-bit input of C_{4.0}″, C_{4.1}″, C_{4.2}″, and C_{4.3}″ one bit position to the right. For example, if C_{4.0}″ is asserted while CARRY is also asserted, then n-channel transistor **474** _{1 }connects node **472**N_{0 }to node **476**N_{1}; that same high signal is then inverted by inverter **480** _{1}, thereby providing an active low signal for C_{4.1}. Similar examples of this right-shifting will be ascertainable by one skilled in the art.

The preceding has demonstrated an electronic processor operable in response to an instruction to efficiently rotate a data argument in response to left and right shift count arguments, without the need to separately shift the entire data argument fully in response to a first shift count followed by then separately shifting the entire data argument fully in response to a second shift count. Given this functionality, additional embodiments of such a processor are now described and achieve additional functions in combination with specific types of left and/or right shift operations. Each of these functions is preferably selected in response to a processor instruction, that is, when a given instruction is issued to processor **100** and when, in response, data rotate unit **200** operates in response to that instruction. Alternatively, the following also may be implemented in still other embodiments using data rotate unit **200**′, but for sake of simplicity the remaining embodiments are shown in connection with data rotate unit **200**. Accordingly, the following discussion first describes various different data manipulation instructions and then is followed by block diagram figures that further augment data rotate unit **200** so as to implement the desired function in response to each such instruction.

A first instruction that may be executed by data rotate unit **200** in one embodiment is referred to herein as an EXTRACT instruction, as that term has heretofore been used in connection with certain DSPs previously sold by Texas Instruments Incorporated. As its name suggests, the EXTRACT instruction operates to extract a number of contiguous bits from a 32-bit data quantity and to place them into a 32-bit result, where the extracted bits are optionally preceded in the result by sign extensions (could instead be zeros) and, as detailed below, in some instances are followed in the result by trailing zeros. Further, the number of bits extracted as well as their positioning in the result are determined in response to the 5-bit left shift argument L and the 5-bit right shift argument R.

Returning to _{1}, where in that example the right shift R is greater than the left shift L and, more particularly, R=10 and L=6. In an effort to appreciate the preferred embodiment, for the EXTRACT instruction note then that in the second shift, that is, the right shift, the resulting MSBs are sign extension bits that duplicate the single MSB from the left-shifted value VL_{1}. In other words, each of those bits matches the MSB after the left shift; thus, in the example illustrated, each sign extension bit matches bit D25, which was the most significant bit in VL_{1 }after the left shift of L=6. Second, the number of sign extension bits equals R bits, as a result of the fact that the right shift of R bits, in effect, shifts in R sign extension bits.

Returning to _{1}, where in that example the right shift R is less than the left shift L and, more particularly, R=8 and L=16. Also in an effort to appreciate the preferred embodiment, some additional observations are now made with respect to _{1 }includes a number of least significant zeros equal to L. Second, in the second shift, that is, the right shift, and when L>R, then a number of these zeros equal to R are shifted out to provide the result VR_{1}. Thus, from the initial data argument, V_{2}, to the result VR,_{1 }and in the case where L>R, then the number of least significant zero bits in the result VR_{1 }equals L−R. Given this latter observation, it also may be stated with respect to the preceding _{1}. In other words, if L−R≦0, then no added least significant zero bits are included in the final result VR_{1}.

Given the preceding observations with respect to the EXTRACT instruction, **200** _{1 }for logically implementing those observations functionally in combination with data rotate unit **200** so as to achieve the desired results for the EXTRACT instruction. System **200** _{1 }includes data rotate unit **200** _{1 }which, as detailed above, rotates its input data argument D[31:0] a distance of R−L, where preferably the equivalent result of such rotation is achieved instead through a rotation of R+LX+1. The 32-bit output from data rotate unit **200** _{1}, DRU[31:0], is connected as a first data input to a 32-bit multiplexer **502**, and adjacent this first input in the schematic of multiplexer **502** is shown a designation of “1”; this designation, as further borne out below, is intended to demonstrate that for each of the 32 control bits, CB[31:0], input to multiplexer **502** and equal to “1”, then multiplexer **502** selects the data at the corresponding bit position of DRU[31:0] and that selected bit is output by multiplexer **502**. For example, assume that the control bit input., CB[31:0], to multiplexer **502** has 10 MSBs equal to 0 (i.e., DRU[31:22]=0), and 22 LSBs equal to 1 (i.e., DRU[21:0]=0). In response to the 22 LSBs equal to 1, multiplexer **502** selects the respective least significant 22 bits of DRU[31:0] output by data rotate unit **200**, that is, it selects DRU[21:0], and it outputs those bits as an input to an AND gate **504**. Note that AND gate **504** is illustrated for simplicity as a single AND gate with two 32-bit inputs. In this regard, one skilled in the art should recognize that AND gate **504** is implemented as 32 AND circuits, where each AND circuit performs an AND operation between a pair of bits that are at a same bit position. Thus, AND gate **404** performs an AND between the MSB at one of its inputs with the MSB at the other of its inputs, and AND gate **504** performs an AND between the bit a position MSB-1 at one of its inputs with the bit at position MSB-1 at the other of its inputs, and so forth for all bits including an AND between the LSB at each of its inputs. The output of AND gate **504** provides the result, V_{EXEC} _{ — } _{EXT}, of executing the EXTRACT instruction.

System **200** _{1 }also includes a sign bit select block **506**, which receives as inputs the values from the left shift argument, L, and the 32-bit data argument D[31:0]. Sign bit select block **506** selects and outputs the bit at position 31−L in the 32-bit data argument D[31:0]. This operation is for providing a sign bit that can fill one or more of the most significant bit positions in the EXTRACT instruction result because, as observed above, those positions are filled with the sign bit that is the most significant bit in the left-shifted result, VL_{1}. In connection with block **506**, therefore, it is recognized that this same bit is at position 31−L in the original data argument D[31:0] and, therefore, block **506** selects that bit to be output. For example, returning to **506** selects the bit at position 31−L=31−6=D25, which can be seen is correctly the MSB of the left-shifted value, VL_{1}. The selected bit is then is output and connected as a second data input to the 32-bit multiplexer **502**, and adjacent this second input in the schematic of multiplexer **502** is shown a designation of “0”; this designation, similar to but in complementary fashion to the “1” designation described above, is intended to demonstrate that for each of the 32 control bits, CB[31:0], input to multiplexer **502** and equal to “0”, then multiplexer **502** selects the sign bit provided by block **506** and outputs it at the corresponding bit position of its output. For example, assume again that the control input to multiplexer **502** has 10 most significant bits equal to 0, and 22 least significant bits equal to 1. In response to the 10 most significant bits equal to 0, multiplexer **502** selects the single sign bit from block **506** and outputs it as the 10 MSBs in its output to AND gate **504**. As further demonstrated below, this output therefore provides the sign extension bit in the most significant bit(s) of the result, V_{EXEC} _{ — } _{EXT}, that is provided from executing the EXTRACT instruction.

System **200** _{1 }also includes an MSB(s) alternate data mask block **508**, which is so named because, as detailed later, causes multiplexer **502** to select alternative data, other than from DRU[31:0], to form part of the result, V_{EXEC} _{ — } _{EXT}. MSB(s) alternate data mask block **508** receives as an input the value of the right shift argument, R, and it outputs a 32-bit thermometer code as a control bit signal, CB[31:0], to multiplexer **502**. As known in the thermometer code art, a thermometer code is a code that includes one set of contiguous zeros and one set of contiguous ones, where the term thermometer is used because of the analogy that the number of one binary state can represent a falling or rising magnitude as in the case of a thermometer. With respect to block **508**, its thermometer code has the number R of its MSBs equal to a zero while any one or more of its LSBs are equal to a one. Further, as introduced above with respect to data rotate unit **200** and sign bit select block **506** and as now appreciated further, the thermometer code provides control bits CB[31:0] and, therefore, causes multiplexer **502** to select a respective bit from either the output of data rotate unit **200** or the output of sign bit select block **506**, and each selected bit is output to AND gate **504**. For example, recall again the example wherein the control bit input, CB[31:0], to multiplexer **502** has 10 MSBs equal to 0, then this control bit input, CB[31:0], is provided as a 32-bit thermometer code from block **508** such as in the form shown in the following Table 7:

TABLE 7 | ||||||||||||||||

Bit | 31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 |

0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 1 | 1 | 1 | 1 | 1 | |

Bit | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |

1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | |

System **200** _{1 }also includes an LSB(s) zero mask block **510**, which is so named because block **510** effectively provides a mask for selecting the zero bits, if any, in the LSB(s) for the final extract instruction result, V_{EXEC} _{ — } _{EXT}. LSB(s) zero mask block **510** receives as inputs the values of the right and left shift argument, R and L, respectively. Block **510** then determines the difference L−R, and if that difference is negative, then block **510** outputs a 32-bit value with all ones. However, if the difference of L−R is positive, then block **510** outputs a 32-bit thermometer code with a number, equal to L−R, of LSB positions having a value zero, while any remaining MSB positions in the code have a value of one. **510**. In either event, the output of block **510** is provided as a second input of AND gate **504**.

The operation of system **200** _{1 }is now described, and is first shown with respect to executing the EXTRACT instruction for the parameters illustrated in _{EXEC} _{ — } _{EXT}, should match that shown as VR_{1 }in _{1}. Looking first to sign bit select block **506**, it selects bit 31−L=31−6, that is, bit **25** (i.e., D25) of the input data D[31:0] and outputs this single bit to the “0” input of multiplexer **502**. Note from **506** has successfully selected the MSB of the left shifted value, VL_{1 }and provided it as an input to multiplexer **502**. Further, MSB(s) alternate data bit mask block **508** provides a thermometer code having R=10 MSBs equal to zeros and the remaining 32−R=32−10=22 LSB positions equal to ones. Accordingly, for this example, block **508** outputs the thermometer code of Table 7 as the control input to multiplexer **502**, thereby providing a value of 0 in the 10 MSBs of that code. In response, multiplexer **502** selects the sign bit data as its “0” input to be the corresponding 10 MSBs in its output, and multiplexer **502** selects the 22 LSBs of DRU[31:0] at its “1” input to be the corresponding LSBs in its output. Accordingly, at this point in its operation, the output of multiplexer **502** is as shown by the preliminary value VPRE_{1 }to the left in **510** outputs a thermometer code with a number of LSBs equal to zero if the difference of L−R is positive; however, in the present example of L=6 and R=10, that difference is negative. In response to the negative difference, instead block **510** outputs a mask of 32 values of one as an input to AND gate **504**, where this mask is shown in _{1}. As known in the art, such a mask when provided as an input to a two-input AND gate causes the gate to simply pass its other input, which in the example of _{PRE1}. Thus, to the right of _{EXEC} _{ — } _{EXT1}, of the executed EXTRACT instruction. Accordingly, one skilled in the art may readily confirm that V_{EXEC} _{ — } _{EXT1 }matches the desired result VR_{1 }of _{EXEC} _{ — } _{EXT1}.

The operation of system **200** _{1 }is now further described by discussing its execution of the EXTRACT instruction for the parameters illustrated in _{2 }in _{2}. Looking to sign bit select block **506**, it selects bit 31−L=31−16, that is, bit **15** (i.e., D15) of the input data D[31:0] and outputs this single bit to the “0” input of multiplexer **502**. Note from **506** has successfully selected the MSB of the left shifted value, VL_{2 }and provided it as an input to multiplexer **502**. Further, MSB(s) alternate data bit mask block **508** provides a thermometer code having R=8 MSBs equal to zeros and the remaining 32−R=32−8=24 LSBs equal to ones. Accordingly, for this example, block **508** outputs the thermometer code of the following Table 8 as the control bit input, CB[31:0], to multiplexer **502**.

TABLE 8 | ||||||||||||||||

Bit | 31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 |

0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | |

Bit | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |

1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | |

In response, multiplexer

From the preceding, one skilled in the art may readily confirm that system **200** _{1 }incorporates the rotation functionality of data rotate unit **200** into a system that provides the desired results for the EXTRACT instruction, given D[31:0] as well as the values of the left and right shift arguments, L and R, respectively. In still other embodiments, however, the present inventive scope may be further modified so as to accommodate still additional data manipulation instructions, as further explored below.

System **200** _{1 }may readily implement a SHIFT RIGHT instruction, where that instruction, as its name suggests, shifts the 32-bit data D[31:0] only to the right and where the extent of the shift is provided by the right shift argument, R. In this case, when the shift right instruction is to be executed, the extent of the left shift can be set to zero, by any of several techniques, such as by providing a hard-wired value of 0 for the left shift for this instruction. In response to the value of L=0, then LSB(s) zero mask block **510** will realize a negative value for L−R and, thus, it will provide a thermometer code of all ones, while at the same time MSB(s) alternate data mask block **508** will provide a thermometer code as described above, that is, with a value of 0 in the R MSB(s) combined with a value of 1 in the remaining LSB(s).

As an additional observation with respect to using system **200** _{1 }to execute a SHIFT RIGHT instruction, note that in some processors the right and left shift arguments are sometimes provided on physical data pins as concatenated values in a single combined argument, as is now also explored in connection with _{ARG1 }that includes a 5-bit right shift pin argument R_{ARG1 }and a 5-bit left shift pin argument L_{ARG1}. In the preferred embodiment, concatenated pin argument C_{ARG1 }represents the actual physical pins that are input to data rotate unit **200** and, therefore, those pins provide to data rotate unit **200** the values of the shift arguments L and R for the EXTRACT instruction described earlier. However, _{ARG2 }that includes a single 6-bit right shift argument R_{ARG2}. In some processors, this one bit larger arguments may be used as the pin inputs for the single argument instructions SHIFT RIGHT and SHIFT LEFT arguments, where any bits in more significant bit positions higher than the argument may be some constant value; in any event, the 6-bit argument provides one more bit in its shift indication as compared to the 5-bit arguments in concatenated argument C_{ARG1 }for the EXTRACT instruction. Note further, therefore, that in a bit-by-bit comparison, bit L_{0 }of concatenated argument C_{ARG1 }aligns with bit R_{5 }of concatenated argument C_{ARG2}. This alignment may need to be considered when implementing system **200** _{1 }to accommodate both the EXTRACT instruction as well as a SHIFT RIGHT instruction. In other words, if data rotate unit **200** were to receive second concatenated pin argument C_{ARG2 }with its 6-bit shift arguments along the same pins as it receives first pin concatenated argument C_{ARG1 }with its 5-bit shift arguments, then without further functionality, a set bit for R_{5 }in C_{ARG2 }may be erroneously interpreted as a set bit for L_{0 }in C_{ARG1}. To address this contingency, in one approach an additional circuit may be provided to override the output of LSB(s) zero mask block **510** by providing additional circuitry between its output and AND gate **504**, where as shown below such circuitry also may provide other utility in connection with other instructions. In another approach, the sixth shift count bit could be provided on a different input pin that is not shared with the left argument. In either case, note that if the most significant bit in the 6-bit right shift argument is set for a SHIFT RIGHT instruction, then the desired right shift is at least 32, thereby shifting out all of the data D[31:0] from the desired result; accordingly, the output of MSB(s) alternate data mask block **508** must be modified such that it provides all zeros in its thermometer code to multiplexer **502**, thereby causing multiplexer **502** to select the sign bit from block **506** and to provide it in all bit positions of its output to multiplexer **502**. Further, because LSB(s) zero mask **510** will necessarily output all ones in its thermometer code for a SHIFT RIGHT instruction (because L−R is negative), then the same output from multiplexer **502** will appear as the ultimate output, V_{EXEC} _{ — } _{EXT}, of the executed instruction.

System **200** _{1 }of **200** _{2}. Before discussing system **200** _{2 }of **200** _{2 }as used for a SHIFT LEFT instruction. To the left of _{PRE3}, and assume that the desired operation of system **200** _{2 }is to shift V_{PRE3 }to the left 16 times. Thus, D[31:0] is provided to data rotate unit **200**, and the value of a left shift argument, L, of 16 is provided to system **200** _{2}. In response to an intermediate function described below, data rotate unit **200** provides a data rotated value V_{DR1}, which as shown in **200** in the preferred embodiment actually performs right rotations; thus, the same value, V_{DR1}, may be described as a right rotation of −L, that is, of −16. Once value V_{DR1 }is provided, then system **200** _{2 }also provides a mask MK_{3}, where the generation of this mask is in connection with various blocks detailed below, and mask MK_{3 }is then logically ANDed with value V_{DR1}. This operation provides the result, V_{EXEC} _{ — } _{SL}, shown to the right of _{EXEC} _{ — } _{SL }matches the desired result of a left shift of V_{PRE3}, leaving a binary value of zero in the 16 LSBs of V_{EXEC} _{ — } _{SL}. Thus, in contrasting the left-shifted result V_{EXEC} _{ — } _{SL }of _{EXEC} _{ — } _{EX2 }of _{EXEC} _{ — } _{SL }includes a zero in one or more LSB positions, while V_{EXEC} _{ — } _{EXT2 }includes a sign extension bit in one or more MSB positions—this difference gives rise to a masking methodology of system **200** _{2 }that contrasts with that of system **200** _{1}, as further explored below.

Looking now to system **200** _{2 }in **200** _{1}, with the following additional modifications. First, a “0” data input is shown to sign bit select block **506**, which replaces the input to that block of data D[31:0]. Second, the output of MSB(s) alternate data bit mask block **508** is no longer directly connected as the control input to multiplexer **502**, but instead that output is connected as an input to a mask modification block **520**, which has its output connected to provide the control bit input, CB[31:0], to multiplexer **502**. The 32-bit output of LSB(s) zero mask block **510** is connected as a first input to an OR gate **522** that provides a 32-bit output connected as an input to AND gate **504**, and the second 32-bit input of OR gate **522** is connected to an OVERRIDE signal. The output of AND gate **504** provides the result of the executed SHIFT LEFT instruction, where that result is indicated in _{EXEC} _{ — } _{SL}. The additional functionality of these various differences as well as newly-added mask modification block **520** is appreciated from the operational description, below. Lastly, for the SHIFT LEFT instruction, only the left shift argument need be provided to system **200** _{2}; however, for reasons detailed below, that argument is preferably manipulated first by a shift count forwarding multiplexer **524**, which then outputs the manipulated value as also explained below.

The operation of system **200** _{2 }to achieve a left shift, in response to a SHIFT LEFT instruction, is now explored, looking first to an issue relating to the left shift argument and its receipt by system **200** _{2}. Specifically, when the SHIFT LEFT instruction is issued, shift count forwarding multiplexer **524** receives the corresponding left shift argument, L. However, in the preferred embodiment, recall that the physical pins of data rotate unit **200** take the form of either concatenated pin argument C_{ARG1 }or C_{ARG2 }in **524** with the SHIFT LEFT instruction, then without additional steps that argument would be provided to the least significant pins of the concatenated pin argument C_{ARGx}. Note, however, in **200** as a right shift argument. For example, consider the **524**, and without additional operation then that left shift value would be provided to the least significant pin positions of concatenated pin argument C_{ARGx}, thereby being received as a desired right shift argument. To compensate for this connectivity, in the preferred embodiment, shift count forwarding multiplexer **524** takes one step toward effectively converting the specified left shift value into a corresponding right shift value. More particularly, as known in the art, a right shift of a number of bit positions is equal to a negative left shift of that same number of bit positions. For the present example where a left shift of 16 bit positions is desired, then that is equivalent to achieving a right shift of negative 16 bit positions, as also observed above with respect to **524** inverts the value of the received left shift argument, L. In addition, as just stated, to form the twos complement then one also must be added to the inverted value. To achieve this additional aspect, recall that in the preferred embodiment, data rotate unit achieves the Equation 1 shift of R+LX+1, which thereby includes the addition of one as its final addend. Thus, data rotate unit **200** can be used to provide the “+1” aspect, and toward this end then shift count forwarding multiplexer **524** inverts the left data shift argument, L, and provides the inverted result to the right pin argument, R_{ARG1}. The result of the proper shifting in response is further appreciated from the remaining discussion, below.

In addition to the preceding, for a SHIFT LEFT instruction, shift count forwarding multiplexer **524** also provides the 5-bit value of L=11111 to the left pin argument, L_{ARG11 }(or L=111111 if a 6-bit, rather than 5-bit, input is used). Particularly, because data rotate unit **200** is to achieve the Equation 1 shift of R+LX+1, then it will therefore require the value of LX, that is, the complement of L. To facilitate the proper result for a SHIFT LEFT instruction, therefore, when shift count forwarding multiplexer **524** provides the value of L=11111 to the left pin argument, L_{ARG11}, then by definition the value of LX is LX=00000. Thus, when data rotate unit **200** in response performs a shift of R+LX+1 positions, then for the SHIFT LEFT instruction it is shifting an extent of R+0+1=R+1; as demonstrated above, R is derived for the SHIFT LEFT instruction as the complement of L, so the result of R+1 provides the twos complement of L and thereby results in a right shift a number of positions equal to that twos complement.

Before detailing additional aspects of the operation of system **200** _{2 }in connection with the SHIFT LEFT instruction, the numeric example of **200** is involved, where recall in that example a left shift of 16 bit positions is desired of V_{PRE3}. The SHIFT LEFT instruction is issued to the processor that includes system **200** _{2}, and at the same time the left shift argument, L=16, is provided to shift count forwarding multiplexer **524**. In binary form, therefore, L=10000 and shift count forwarding multiplexer **524** determines its complement, 01111, and provides that complement value to the 5 least significant bits of concatenated pin argument C_{ARG1 }(or 6 bits to concatenated pin argument C_{ARG2}); thus, these bits are received by data rotate unit **200** as the right shift pin argument, R_{ARG1}. Additionally, in response to the SHIFT LEFT instruction, shift count forwarding multiplexer **524** provides the value L=11111 to the left shift pin argument, L_{ARG11}, such that its complement, L_{XARG1}, is 00000. Next, data rotate unit **200** performs a right rotation a number of bit positions equal to R+LX+1, which in this example is 01111+00000+1, which equals a binary value of 10000 or a decimal value of 16. Accordingly, in this example, the desired result for a left shift of 16 is reached by effectively rotating right by 16 positions. Further, additional masking is also performed by system **200** _{2 }so as to achieve the final result, V_{EXEC} _{ — } _{SL}, of

Continuing now with additional operations of system **200** _{2}, recall that LSB(s) zero mask block **510** determines the difference L−R and outputs a responsive thermometer code; further, from the preceding discussion, for the SHIFT LEFT instruction, L_{ARG1}=11111 and, thus, the resulting thermometer code would not provide a proper mask to use in connection with the output of multiplexer **502**. In this regard, instead, when the SHIFT LEFT instruction is executed, then OVERRIDE is asserted high, which provides 32 high inputs to OR gate **522** and that are OR'd with each output from block **510**, thereby effectively overriding them in that the entire 32-bit output of OR gate **522** will be high. Moreover, because these high 32 bits are provided as inputs to AND gate **504**, then they will cause the output of AND gate **504** to provide whatever values are received at its other input from the output of multiplexer **502**. From the preceding, therefore, one skilled in the art will recognize that asserting the OVERRIDE signal effectively eliminates the output of block **510** from affecting the result to be produced by the SHIFT LEFT instruction.

Completing the discussion of the operation of system **200** _{2}, mask modification block **520** is provided to achieve the proper masking to include any necessary zero values in the LSB(s) of the final result of the shift operation. For example with reference to **520** from _{3 }so that the 16 LSBs in V_{EXEC} _{ — } _{SL }are provided with a value of zero. First, recall that MSB(s) alternate data mask block **508** receives a right shift argument, R, and it outputs a thermometer code with R most significant bit position(s) equal to a value of 0; for the EXTRACT and SHIFT RIGHT instructions those 0's thereby select the output of sign bit select block **506** to appear in the MSB(s) of the result. In contrast, for system **200** _{2 }to achieve a SHIFT LEFT instruction, it is desired for the LSB(s) of the result to include a value of zero. Toward this end, a value of zero is input as an input to sign bit select block **506** (or is wired into that circuit for use in response to the SHIFT LEFT instruction), so that value may be selected by multiplexer **502** for the appropriate LSB(s) in the instruction result. Since the zero value is for the LSB(s) rather than the MSB(s), then the thermometer code control bit input, CB[31:0], to multiplexer **502** must be modified, as is achieved by mask modification block **520**. Particularly, in the preferred embodiment, mask modification block **520** receives the thermometer code from MSB(s) alternate data mask block **508**, inverts that thermometer code and shifts it right by one bit, thereby outputting the resultant modified thermometer code as CB[31:0] to control multiplexer **502**. Again, consider the example of **524** determines its complement, 01111, and provides that as the right shift pin argument, R_{ARG1}, to data rotate unit **200**. Thus, MSB(s) alternate data mask block **508** receives the binary value of R_{ARG1}=01111, which corresponds to a decimal value of 15. Accordingly, MSB(s) alternate data mask block **508** outputs a thermometer code with a zero located in each of its 15 MSBs and, consequently, with a one located in each of its remaining 17 LSBs. However, mask modification block **520** receives this input thermometer code and inverts it, thereby producing an intermediate modified thermometer code with a one located in each of its 15 MSBs and, consequently, with a zero located in each of its remaining 17 LSBs; next, block **520** shifts that intermediate modified thermometer code one position to the right (bringing in a “1” in the MSB of the shifted value), thereby producing a final modified thermometer code, corresponding to mask MK_{3 }in **200** _{2 }is not logically ANDed, but instead it controls multiplexer **502**. Thus, in response to the one located in each of the 16 MSBs of the final modified thermometer code, multiplexer **502** selects and outputs the corresponding bit position value output by data rotate unit **200**, thereby outputting bits D15:D0 in the 16 MSBs of the multiplexer output. At the same time, in response to the zero located in each of the 16 LSBs of the final modified thermometer code, multiplexer **502** selects and outputs the value of zero provided by sign bit select block **506**, thereby outputting a value of zero in each of the 16 LSBs of the multiplexer output. Finally, the output of multiplexer **502** is connected to AND gate **504**, but recall due to the assertion of the OVERRIDE signal concurrent with the execution of the SHIFT LEFT instruction, the output of multiplexer **502** simply passes through AND gate **504**, thereby providing the same result as shown as the result, V_{EXEC} _{ — } _{SL}, shown to the right of **508** and mask modification block **520** must be modified such that code bits CB[31:0] provides all zeros to multiplexer **502**, thereby causing multiplexer **502** to select the value of “0” from block **506** and to provide it in all bit positions of its output to multiplexer **502**.

System **200** _{1 }of **200** _{3}. Before discussing system **200** _{3 }of **200** _{3 }as used for either the SET or CLEAR instruction. In general, the SET instruction sets, and the CLEAR instruction clears, data bit positions between specified top and bottom locations in a data input, where those specified locations are indicated by the values of the shift arguments R and L, respectively. Thus, with respect to _{PRE4}, and assume that the desired operation of system **200** _{3 }is to execute either a SET or CLEAR instruction. For either such instruction the values of the R and L shift arguments indicate a respective top and bottom bit position within V_{PRE4}, where the data in the bit position at the specified bottom bit position and up to and including the top bit position are to be either set to one (for the SET instruction) or cleared to zero (for the CLEAR instruction), and where the remaining data, that is, starting above the top bit position, and below the bottom bit position, remains unchanged from the input data. By way of example, _{EXE} _{ — } _{S/C }by “S/C” in the corresponding bit position. Additionally, the argument of R=21 causes the data D21 at the bit position 21 to be set or cleared, as also indicated in result value V_{EXE} _{ — } _{S/C }by “S/C” in the corresponding bit position. Finally, in addition to setting or clearing the bit at positions 14 and 21, the data between those bit positions, that is, the remaining bits D15 through D20, are also set or cleared in response to whether the instruction is a SET or CLEAR instruction, respectively. Accordingly, result V_{EXE} _{ — } _{S/C }ultimately provides a block BL_{1 }of set/cleared bits, yet outside of those bits the remaining data from V_{PRE4 }is carried forward to the respective bit positions of V_{EXEC} _{ — } _{S/C}.

Having demonstrated the results of the SET and CLEAR instructions, attention is returned to system **200** _{3 }of **200** _{1 }of _{EXEC} _{ — } _{S/C}, for the SET and CLEAR instructions. System **200** _{3 }essentially bypasses the rotating function of data rotate unit **200** and instead may use unit **200** to provide either a one output for the SET instruction or a zero output for a CLEAR instructions, as is now explored. Looking then to system **200** _{3 }in detail, the data D[31:0] is input to a data gate **530**, which gates that data and outputs it in unchanged from to the “0” input of multiplexer **502**, which as described for earlier embodiments therefore selects this data at each bit position corresponding to a “0” in the same bit position of the 32-bit input control bit code, CB[31:0], to multiplexer **502**. The data D[31:0] is also input to data rotate unit **200**, but for reasons appreciated below unit **200**, when executing the SET instruction outputs 32 values of one and when executing the CLEAR instruction outputs 32 zeros; thus, the input data D[31:0] is of no effect for data rotate unit **200** with respect to these two instructions. As an alternative to achieve this same result, **200** is connected as a first input **540** _{1 }to a multiplexer **540**, which has its second input **540** _{2 }connected to receive either a value of 1 for the SET instruction or a value of 0 for the CLEAR instruction. Multiplexer **540** also receives a control signal shown as S&C, which is asserted for both the SET and CLEAR instructions, thereby causing multiplexer **540** to select its second input **540** _{2}, thereby selecting either 1 or 0, respectively, and providing it at each bit position in a 32-bit output. The 5-bit left and right shift arguments, L and R, respectively are again input to a shift count forwarding multiplexer **524**. In response, shift count forwarding multiplexer **524** couples the value of L to a 5-bit left shift pin argument L_{ARG11}; however, with respect to the value of R, shift count forwarding multiplexer **524** couples its complement to the 5-bit right shift pin argument R_{ARG1}, for reasons appreciated below.

System **200** _{3 }includes the MSB(s) alternate data mask block **508** of system **200** _{1 }and which receives the 5-bit right shift pin argument R_{ARG1}, but it does not include LSB(s) zero mask block **510**. Instead, system **200** _{3 }includes a set/clear mask block **550** which in the preferred embodiment receives the 5-bit left shift pin argument L_{ARG11}, and in response block **550** outputs a 32-bit thermometer code that has a value of zero in its L_{ARG1 }LSBs, with a value of one in its remaining MSBs. Thus, consistent with the earlier illustrations, **550**. This code is connected to a mask combine block **560**, which also receives the thermometer code output from MSB(s) alternate data mask block **508**. In the preferred embodiment, mask combine block **560** performs a logical AND with each corresponding bit position from the two 32-bit thermometer codes it receives as inputs, that is, bit **31** of the two codes are ANDed together and the result is provided as bit **31** at output **560** _{out}, while similarly bit **30** of the two codes are ANDed together and the result is provided as bit **30** at output **560** _{out}, and so forth for the remaining respective bits in the two input thermometer codes. The resulting 32-bit code at output **560** _{out }is connected as the control bit code, CB[31:0], to multiplexer **502**.

The operation of system **200** _{3 }is now described, and by way of illustration the example of _{1 }and TC_{2 }and mask MK_{4 }produced by system **200** _{3}, as are now examined. Set/clear mask block **550** receives the value of L_{ARG1}=L=14, and in response it outputs thermometer code TC_{1}, having the value zero in L=14 LSB positions, with a value of one in the remaining 18 MSB positions. MSB(s) alternate data mask block **508** receives R_{ARG1}, which recall for the SET or CLEAR instruction is provided by shift count forwarding multiplexer **524** as the complement of R. Thus, in the present example R=21 which is a binary value of 10101, and thus its complement of 01010, or decimal 10, is provided as R_{ARG1 }to MSB(s) alternate data mask **508**. In response, block **508** outputs thermometer code TC_{2}, having the value zero in R_{ARG2}=10 (decimal) MSB positions, with a value of one in the remaining 22 LSB positions. Next, thermometer codes TC_{1 }and TC_{2 }are connected to mask combine block **508**, which performs a bit-by-bit logical AND of each respective bit position; this operation produces and outputs mask MK_{4 }at output **560** _{out}. One skilled in the art may readily confirm, therefore, that mask MK_{4 }includes a value of one only in the bit positions that correspond to bits [D21:D14] from V_{PRE4}, where the remaining bit positions in mask MK_{4 }include a value of zero. Further, mask MK_{4 }is connected to control multiplexer **502**. Thus, in response to the logical 1 at bit positions [21:14] in mask MK_{4}, multiplexer **502** selects from its “1” input the output from multiplexer **540**, where multiplexer **540** provides either a zero for the CLEAR instruction or a one for the SET instruction. This value is selected and output by multiplexer **502** to bit positions [21:14] in the final executed result, V_{EXEC} _{ — } _{S/C}, which thereby forms block BL_{1}. In block BL_{1}, for sake of illustrating both the SET and CLEAR instructions, its bits are shown to include a set/clear designation “S/C” in the set/cleared bit positions. At all bit positions other than bit positions [21:14], mask MK_{4 }includes a logical zero; hence, for these other bit positions, multiplexer **502** selects from its “0” input the output from data gate **530**, that is, the bits from the respective bit positions in D[31:0]. Thus, for the bit positions in the execution result, V_{EXEC} _{ — } _{S/C}, outside of block BL_{1}, those bit positions, V_{EXEC} _{ — } _{S/C}[13:0] and V_{EXEC} _{ — } _{S/C}[31:22], are equal to the data from the respective bit positions in D[31:0].

System **200** _{1 }of **200** _{4}. Before discussing system **200** _{4 }of _{1}, SH_{2}, and SH_{1}. Starting to the left of _{0 }through A_{3 }and B_{0 }through B_{3}, where the two data arguments are concatenated to form a 64-bit quantity. This same quantity is repeated twice more in _{1}. To obtain SH_{1}, the concatenated 64-bit quantity is shifted one byte to the right, and the least significant 32 bits form the result, SH_{1}. This same general technique is also illustrated for a FUNNEL SHIFT instruction executing a two-byte shift, SH_{2 }in the middle of _{3}, toward the right of _{x }is the 32 LSBs after the right shift.

Having demonstrated the results of the FUNNEL SHIFT instruction, attention is returned to system **200** _{4 }of **200** _{1 }of _{EXEC} _{ — } _{FC}. As further shown below, system **200** _{4 }requires only the right rotation operation of data rotate unit **200** and the selection of its output by multiplexer **502**; thus, to simplify the figure other aspects from previous embodiments are not shown, while two additional aspects are included so as to support the FUNNEL SHIFT instruction. System **200** _{4 }includes a rotate data forwarding multiplexer **570**, which is connected to receive two 32-bit arguments, shown as A[31:0] and B[31:0]. Logically, these two values are shown to be stored in respective storage devices **570** _{SA }and **570** _{SB}, where those devices actually may merely represent the gating of such data as opposed to more complex storage circuitry. For sake of reference, each of storage devices **570** _{SA }and **570** _{SB }is shown to include four separate bytes from A[31:0] and B[31:0], respectively. Rotate data forwarding multiplexer **570** also includes four single-byte multiplexers **570** _{BM0}, **570** _{BM1}, **570** _{BM2}, and **570** _{BM3}, where each such multiplexer has one input connected to receive a byte from storage device **570** _{SA }and another input connected to receive a like-positioned byte from storage device **570** _{SB}. Thus, byte multiplexer **570** _{BM0 }is connected to receive byte A_{0 }and byte B_{0}, byte multiplexer **570** _{BM1 }is connected to receive byte A_{1 }and byte B_{1}, byte multiplexer **570** _{BM2 }is connected to receive byte A_{2 }and byte B_{2}, and byte multiplexer **570** _{BM3 }is connected to receive byte A_{3 }and byte B_{3}. Rotate data forwarding multiplexer **570** also includes a data storage area **570** _{D}, again that may be implemented simply by gating data, and which receives a single byte from the output of each of byte multiplexers **570** _{BM0}, **570** _{BM1}, **570** _{BM2}, and **570** _{BM3}. The combined storage from area **570** _{D }thus provides data D[31:0] to data rotate unit **200**. Data rotate unit **200**, in the same manner as described above, also receives the value of the R shift argument, which may come in the form of a right shift pin argument R_{ARG1 }(not shown); for reasons appreciated below, the value of the left shift argument, L, has no effect on system **200** _{4 }for realizing the FUNNEL SHIFT instruction and, thus, it is not shown in **200** _{4 }also includes an all ones mask adjust block **572**, which in executing the FUNNEL SHIFT instruction outputs a 32-bit code C_{FS }that includes all values of one. Thus, unlike previous systems, no thermometer code having both ones and zeros is required. Indeed, if system **200** _{4 }is combined with other previously-described systems, then any such thermometer codes may be input to all ones mask adjust block **572** which, in response, will in place of such a code(s) output the all-one-value code C_{FS}, where that output is connected to the control input of multiplexer **502**. Moreover, because only ones are provided as a control input to multiplexer **502**, then multiplexer **502** will necessarily select the entire 32 bits from its “1” input, thereby ignoring the value, if any, provided to its “0” input; for simplification, therefore, in **502**, although an input may be provided yet it will be ignored during execution of the FUNNEL SHIFT instruction. Lastly, the output of multiplexer **502** provides the executed result, V_{EXEC} _{ — } _{FS}, for the FUNNEL SHIFT instruction.

The operation of system **200** _{4 }is now described. In general, system **200** _{4 }performs a right rotation of data D[31:0], where due to the byte granularity nature of the FUNNEL SHIFT instruction, the extent of the right shift, as indicated in R or R_{ARG1}, will be a number of bit positions equal to either 8, 16, or 24 bits (i.e., 1, 2, or 3 bytes). Moreover, because code C_{FS }provides all ones to the control input of multiplexer **502**, then the final executed result, V_{EXEC} _{ — } _{FS}, provides the entire rotated value of D[31:0] without any additional masking or changing of bits. To fully appreciate, however, that each such result can match that of the desired shift examples in **570**, as discussed immediately below.

The operation of rotate data forwarding multiplexer **570** and the additional data manipulation of system **200** _{4 }using the FUNNEL SHIFT instruction is illustrated in connection with **570** prior to being right-byte rotated by data rotate unit **200** of system **200** _{4}. As a first example, looking to the left of **570** _{D }where the value is shown as D_{1}[31:0]. Value D_{1}[31:0] is achieved by controlling byte multiplexer **570** _{BM0 }to select B_{0 }from storage area **570** _{SB}, while each of byte multiplexers **570** _{BM1}, **570** _{BM2}, and **570** _{BM3 }select respective bytes A_{1}, A_{2}, and A_{3 }from storage area **570** _{SA}. Immediately to the right of value D_{1}[31:0], **200**, while value D1[31:0] is provided as data. The right rotation provides the result SH_{1}, which due to the all one values in code C_{FS}, passes through multiplexer **502** to provided V_{EXEC} _{ — } _{FS}. Now, comparing _{1}, is realized in both figures. Thus, _{1 }may be achieved by first manipulating the data arguments A[31:0] and B[31:0] using rotate data forwarding multiplexer **570**, followed by a right byte rotation using data rotate unit **200**.

The remainder of _{2}[31:0] demonstrates the instance where byte multiplexers **570** _{BM0 }and **570** _{BM1 }select respective bytes B_{0 }and B_{1 }from storage area **570** _{SB}, while byte multiplexers **570** _{BM2 }and **570** _{BM3 }select respective bytes A_{2 }and A_{3 }from storage area **570** _{SA}. Immediately to the right of value D_{2}[31:0], _{2}, is provided as V_{EXEC} _{ — } _{FS}. Also, again comparing _{2}, is realized in both figures. Completing _{3}[31:0] demonstrates the instance where byte multiplexers **570** _{BM0}, **570** _{BM1}, and **570** _{BM2 }select respective bytes B_{0}, B_{1}, and B_{2 }from storage area **570** _{SB}, while byte multiplexer **570** _{BM3 }selects byte A_{3 }from storage area **570** _{SA}. Immediately to the right of value D_{3}[31:0], _{3}, is provided as V_{EXEC} _{ — } _{FS}. Once more comparing _{3}, is realized in both figures.

System **200** _{1 }of **200** _{2 }of **200** _{5}, shown in **200** _{5 }includes many of the same aspects of system **200** _{2}, and the reader is assumed familiar with those aspects due to the earlier explanation of them in connection with system **200** _{2}. Where system **200** _{5 }differs from system **200** _{2}, is in the addition of a saturation detect block **580** that is connected to receive both the data to be rotated, D[31:0], as well as the 5-bit right shift pin argument R_{ARG1}, which recall for a SHIFT LEFT instruction actually provides the value of the desired left shift argument, L, as received on the five least significant pins of the concatenated pin argument C_{ARG1}. Saturation detect block **580** provides a control signal to a saturation constant switch **582**. Saturation constant switch **582** is also connected to receive the left-shifted data, D_{LS}[31:0], from the output of multiplexer **502**, and to output a corresponding and potentially modified value, D_{M}[31:0], to an input of AND gate **504**.

The operation of system **200** _{5 }to execute a SATURATING SHIFT LEFT instruction is now explored, first by looking to saturation detect block **580**. Saturation detect block **580** determines from both D[31:0] and the desired amount of left shift (as provided by R_{ARG1}) whether the result following the desired shift will exceed the 32-bit precision of system **200** _{5}. Such a device may be constructed in various manners by one skilled in the art. In the preferred embodiment, block **580** determines whether the most significant digit (“MSD”) in D[31:0], if shifted left the desired amount, would reach the MSB position (i.e., bit position 31) of data rotate unit **200** or higher. Note here that the term “most significant digit” depends on whether the data D[31:0] is positive or negative; if positive, then the MSD is the most significant bit position containing a value of one, whereas if negative then the MSD is the most significant bit position containing a value of zero. If the desired shift will bring that MSD into bit position 31, or higher, of data rotate unit **200**, then that shift will violate the requirement that the MSB position be reserved to provide a sign bit, thereby exceeding the precision of data rotate unit **200**. In this case, saturation detect block **580** asserts its control signal SAT_CTRL, to saturation constant switch **580**, where in contrast if saturation would not occur after the desired shift, then saturation detect block **580** does not assert SAT_CTRL.

In general, saturation constant switch **582** either passes its input directly to its output when no saturation is occurring, or alternatively it outputs the appropriate saturation constant when saturation is occurring. More specifically, when SAT_CTRL is asserted, then switch **582** is informed that saturation is occurring. In response, it outputs either a positive or negative saturation constant, based respectively on whether the data to be rotated, D[31:0], is positive or negative. If D[31:0] is positive, then for a 32-bit number, the positive saturation constant is 0×7fffffff, that is, the largest magnitude positive number permissible for 32-bit precision. If D[31:0] is negative, then for a 32-bit number, the negative saturation constant is 0×80000000, that is, the largest magnitude negative number permissible for 32-bit precision. Thus, when positive saturation exists, then D_{M}[31:0]=0×7fffffff, when negative saturation exists, then D_{M}[31:0]=0×80000000, and if no saturation exists, then D_{M}[31:0]=D_{LS}[31:0] (i.e., the input passes unchanged to the output). In any event, the remainder of system **200** _{5 }operates in the same manner as system **200** _{2 }in connection with the SHIFT LEFT instruction. Thus, once D_{M}[31:0] is provided, and recall that OVERRIDE is asserted high for the SHIFT LEFT instruction, then D_{M}[31:0] is passed by AND gate **504** to provide the result, V_{EXEC} _{ — } _{SSL}, for having executed the SATURATING SHIFT LEFT instruction.

**582** of system **200** _{5}. By way of introduction, note that the functionality of alternatively permitting D_{LS}[31:0] to either pass through switch **582** or be replaced with a saturation constant could be implemented using a multiplexer. However, such an approach may be undesirably slow in some applications. Accordingly, in the preferred embodiment, switch **582** is implemented using logic. More particularly, **580** when positive or negative saturation, respectively, occurs, and consequently the complements, satplusx and satminusx are asserted by saturation detect block **580** when positive or negative saturation, respectively, does not occur. The most significant bit, D_{LS}[31], of the input D_{LS}[31:0] to switch **528**, is connected as a first input to an AND gate **582** _{AN1}, which has its output connected to a first input of an OR gate **582** _{OR1}, where gates **582** _{AN1}, and **582** _{OR1 }are preferably formed as a single complex OR-after-AND gate. A second input of AND gate **582** _{AN1 }is connected to receive the satplusx signal, and a second input of OR gate **582** _{OR1 }is connected to receive the satminus signal. The remaining 31 LSBs, D_{LS}[30:0], of D_{LS}[31:0], are connected to a same type of circuit, one of which is shown by way of example in _{LS}[30:0]. Specifically, each of those bits is connected as a first input to an AND gate **528** _{AN2}, which has its output connected to a first input of an OR gate **582** _{OR2}, where gates **582** _{AN2 }and **582** _{OR2 }are preferably formed as a single complex OR-after-AND gate. A second input of AND gate **582** _{AN2 }is connected to receive the satminusx signal, and a second input of OR gate **582** _{OR2 }is connected to receive the satplus signal.

The operation of switch **582** is now described. In general, when saturation is detected, switch **582**, using logic circuitry, effectively transforms the input to the appropriate one of either the positive or negative saturation constant, whereas when saturation is not detected, switch **582**, using the same logic circuitry, allows the input D_{LS}[31:0] to pass to the output and become D_{M}[31:0]. Each of these instances is now explored in greater detail. By way of further appreciation of switch **582**, the following Table 9 provides the various alternative states of the control signal in switch **582** based on whether saturation occurs and, if so, whether that saturation is positive or negative.

TABLE 9 | |||

Positive | Negative | No saturation | |

Signal | saturation value | saturation value | value |

satplus | 1 | 0 | 0 |

satplusx | 0 | 1 | 1 |

satminus | 0 | 1 | 0 |

satminusx | 1 | 0 | 1 |

Looking at the operation of switch **582** when positive saturation is detected, then the four control signals in SAT_CTRL as provided by saturation detect block **580** are as shown in the first value column of Table 9. In response to satplus=1, the output of OR gate **582** _{OR2 }equals one regardless of the output of AND gate **582** _{AN2 }and, thus, each of the bits in D_{M}[30:0] equals 1. At the same time, in response to satplusx=0, the output of AND gate **582** _{AN1 }equals 0, while at the same time satminus=0; thus, the output of OR gate **582** _{OR1}, that is, bit D_{M}[31], equals 0. Accordingly, when positive saturation is detected, D_{M}[31:0] equals the desired positive saturation constant of 0×7fffffff.

Looking at the operation of switch **582** when negative saturation is detected, then the four control signals in SAT_CTRL as provided by saturation detect block **580** are as shown in the second value column of Table 9. In response to satminus=1, the output of OR gate **582** _{OR 1 }equals one regardless of the output of AND gate **582** _{AN1 }and, thus, bit D_{M}[31], equals 1. At the same time, in response to satminusx=0, the output of AND gate **582** _{AN2 }equals 0, while at the same time satplus=0; thus, the output of OR gate **582** _{OR2 }equals 0, and, hence, each of the bits in D_{M}[30:0] equals 0. Accordingly, when negative saturation is detected, D_{M}[31:0] equals the desired negative saturation constant of 0×80000000.

Looking at the operation of switch **582** when no saturation is detected, then the four control signals in SAT_CTRL as provided by saturation detect block **580** are as shown in the third value column of Table 9. In response to satplusx=1, the output of AND gate **582** _{AN1}, passes the data D_{LS}[31], and in response to satminus=0, the output of OR gate **582** _{OR1 }passes the output of AND gate **582** _{AN1}, which is passing D_{LS}[31]. Thus, when no saturation is detected, D_{M}[31] equals D_{LS}[31]. Additionally when no saturation is detected, in response to satminusx=1, the output of AND gate **582** _{AN2 }passes the data D_{LS}[30:0], and in response to satplus=0, the output of OR gate **582** _{OR2 }passes the output of AND gate **582** _{AN2}, which is passing D_{LS}[30:0]. Thus, when no saturation is detected, D_{M}[30:0] equals D_{LS}[30:0]. The preceding therefore demonstrates that again the desired result is reached, that is, when no saturation is detected, D_{M}[31:0] equals D_{LS}[31:0].

Having described various systems **200** _{1}, through **200** _{5}, **200** _{6 }with a common data rotate unit **200** to achieve all of the above-described instructions, namely: EXTRACT, SHIFT RIGHT, SHIFT LEFT, SET, CLEAR, FUNNEL SHIFT, and SATURATING SHIFT LEFT. From comparing **520** of system **200** _{2 }in **560** of system **200** _{3 }in **572** of system **200** _{4 }in **508** and set/clear mask block **550**, then **590**, where that block is therefore intended to encompass all of the above-described potential mask adjustments, and in which case the adjusted value is used to provide the control bit signal CB[31:0], to control multiplexer **502**. Moreover, also in connection with multiplexer **502**, the various alternative data inputs are shown with respect to its “0” input, such as the possibility of receiving the sign bit or a value of zero from sign bit select block **506**, or the input data D[31:0] from data gate **530**. Also in connection with multiplexer **502** are shown the various alternative data inputs with respect to its “1” input, including the output from data rotate unit **200** or a set or clear bit, the selection of these being controlled by multiplexer **540**. Lastly, the output of multiplexer **502** may pass directly through, or be adjusted by, saturation constant switch **582**, and is then ANDed via AND gate **504** with the output of LSB(s) zero mask block **510**, unless that block's output is overridden by assertion of the OVERRIDE signal to OR gate **522**, in which case AND gate **504** passes its other input to its output. Finally, therefore, the output of AND gate **504** produces the resulting value, V_{EXEC} _{ — } _{F}, corresponding to the execution of whatever one of the above-described selected instructions.

**590** of **502** for various instructions. By way of introduction, recall that mask adjust block **590** combines the functions of various of the mask adjusting blocks described above. Block **590** includes an OR gate **590** _{OR1 }having one input for receiving a signal shlx, which as an active low signal is low when the SHIFT LEFT instruction is executed, and OR gate **590** _{OR1 }includes another input for receiving shiftleft_maskx, the active low mask corresponding to the SHIFT LEFT instruction. The output of OR gate **590** _{OR1 }is connected as one input to an NAND gate **590** _{NAN1}, where OR gate **590** _{OR1 }and NAND gate **590** _{NAN1 }are preferably constructed as a single complex gate. Another input of NAND gate **590** _{NAN1}, receives the signal fshx, which is an active low signal asserted low for execution of the FUNNEL SHIFT instruction. The output of NAND gate **590** _{NAN1 }is connected as an input to a NOR gate **590** _{NOR1}, which has its other input connected to the output of an AND gate **590** _{AN1}. One input of AND gate **590** _{AN1 }receives the shiftright_mask signal, which represents a bit from the mask from MSB(s) alternate data mask block **508** and where shiftleft_maskx is created by downshifting shiftright_mask one bit. The other input of AND gate **590** _{AN1 }receives the output of a NAND gate **590** _{NAN2}. One input of NAND gate **590** _{NAN2 }is connected to the output of an OR gate **590** _{OR2}, and the other input of NAND gate **590** _{NAN2 }is connected to receive a signal shrx, an active low signal asserted low for the SHIFT RIGHT instruction. One input of OR gate **590** _{OR2 }is connected to receive the signal setclrx, which is an active low signal asserted low for execution of either the SET or CLEAR instruction. The other input of OR gate **590** _{OR2 }is connected to receive the setclr_mask, which represents a bit from the set/clear mask block **550**. Lastly, the output of NOR gate **590** _{NOR1 }is connected through an inverter **590** _{INV}, and the output of inverter **590** _{INV }provides the control bit signal CB[31:0].

The operation of the schematic of **590** should be readily understood given the earlier description of the block functionality as well as the signal definitions provided with respect to

With respect to **590** _{NAN2 }to be high and this high is connected as an input to AND gate **590** _{AN1}, which causes AND gate **590** _{AN1 }to pass to its output the value at its other input, shiftright_mask; at the same time, shlx is high which forces a high output out of OR gate **590** _{OR1}, and that high is combined with the high of fshx causing the output of NAND gate **590** _{NAN1 }to be low. This low is connected as an input to NOR gate **590** _{NOR1}, which as a result outputs the complement of its other input, which as shown above is the shift right_mask as passed by AND gate **590** _{AN1}. This complement is then inverted by inverter **590** _{INV}, thereby passing the shift right_mask bit to multiplexer **502**.

With respect to **590** _{CR1 }to output the value of its other input, namely, shiftleft_maskx, that is, the complement of the left shift mask bit. At the same time, fshx is high, thereby causing NAND gate **590** _{NAN1 }to pass to its output its other input, namely, the value of shiftleft_maskx and, further, NAND gate **590** _{NAN1 }forms the complement of that data, thereby providing the actual value shiftleft_mask as an input to NOR gate **590** _{NOR1}. At the same time, setclrx is high causing OR gate **590** _{OR2 }to output a high as one input to NAND gate **590** _{NAN2}, and the other input to that NAND gate **590** _{NAN2}, shrx, is also high, thereby causing NAND gate **590** _{NAN2 }to output a low to AND gate **590** _{AN1 }and, hence, also causing that AND gate **590** _{AN1 }also to output a low to NOR gate **590** _{NOR1}. This low received by NOR gate **590** _{NOR1 }causes it to pass to its output the value at its other input, which recall has been shown to be the actual value shiftleft_mask, and as a NOR gate, that value is complemented when it is ouput. However, the complement, shiftleft_maskx, then is inverted again as it passes through inverter **590** _{INV}, in which case inverter **590** _{INv }then outputs the value of shiftleft_mask to multiplexer **502**.

With respect to **590** _{OR2 }to output the value of its other input, namely, setclr_maskx, that is, the complement of the set/clear mask bit. At the same time, shrx is high, causing NAND gate **590** _{NAN2 }to output the complement of its other input, where that input has been shown to be setclr_maskx; thus, NAND gate **590** _{NAN2 }outputs setclr_mask to AND gate **590** _{AN1}, which is receiving shiftright_mask at its other input. Also at the same time, shlx is high thereby forcing the output of OR gate **590** _{OR1 }high, and this high input is provided to NAND gate **590** _{NAN1}, along with another high input, fshx; thus, the output of NAND gate **590** _{NAN1 }is forced low by these two high inputs, and this low is connected to NOR gate **590** _{NOR1}, thereby causing it to output the complement of its other input. That other input has been shown above to be setclr_mask, which is inverted by NOR gate **590** _{NOR1 }and, thus, output as setclr_maskx. However, that inverted value is then inverted again by inverter **590** _{INV}, thereby outputting the value of setclr_mask ANDed with shiftright_mask to multiplexer **502**.

With respect to **590** _{NAN1 }to be high. This high is input to NOR gate **590** _{NOR1}, thereby forcing its output to be low, and that low is then inverted by inverter **590** _{INV }to output a one to multiplexer **502**. Thus, for each of the 32 bits in CB[31:0], then each circuit similar to the schematic of **572**, as used for the FUNNEL SHIFT instruction.

**600** for implementing various of the signal selectivity provided by multiplexer **502**, saturation constant switch **582**, and AND gate **504** of _{EXEC} _{ — } _{F}; thus, system **600** is in the preferred embodiment duplicated for a total of 32 like-design circuits, one for each bit in V_{EXEC} _{ — } _{F}. Turning to the signals and devices in circuit **600**, a signal sign_bitx, representing an active low output of sign bit select block **506**, is connected as an input to an AND gate **600** _{AN1}. Another input of AND gate **600** _{AN1 }is connected to receive a signal data_gatex, an active low signal representing the output of data gate **530**. The output of AND gate **600** _{AN1 }is connected as an input to a NOR gate **600** _{NO1}, which has another input connected to receive a signal mask_adj, which is one bit of the control bit code, CB[31:0], output by mask adjust block **590**. The signal mask_adj is also connected as an input to an AND gate **600** _{AN2}, which also receives a signal, rot_data, representing a single bit of the output signal, DRU[31:0], from data rotate unit **200**. The output of AND gate **600** _{AN2 }is connected as an input to an OR gate **600** _{OR1}, which receives as another input the output of NOR gate **600** _{NO1}. The output of OR gate **600** _{OR1 }is connected as an input to an NAND gate **600** _{NAN1}. Another input of NAND gate **600** _{NAN1 }is connected to the output of an OR gate **600** _{OR2}. One input of OR gate **600** _{OR2 }is connected to receive a signal satx_{—}0, which is asserted active low when a saturation output bit of 0 is desired, and another input of OR gate **600** _{OR2 }is connected to a signal, lsbz_mask, representing one bit of the output of LSB(s) zero mask block **510**. The output of NAND gate **600** _{NAN1 }is connected as an input to a NAND gate **600** _{NAN2}, which has a another input connected to receive a signal satx_{—}1, which is asserted active low asserted when a saturation output bit of 1 is desired.

The operation of circuit system **600** will be understood by one skilled in the art given the above-described signal definitions as well as the functional descriptions provided earlier and, thus, only certain aspects are briefly observed here. First, note that the signal lsbz_mask is valid only for an EXTRACT instruction and, for all other instructions, it is a value of zero (comparable to the complement of the effect of the above-described OVERRIDE signal). Next, looking to the instances of saturation, when it is desired for the output bit of V_{EXEC} _{ — } _{F }to be a 1 (e.g., for the MSB in negative saturation or for all bits other than the MSB in positive saturation), then satx_{—}1 is asserted low and satx_{—}0 is high. The former forces the output V_{EXEC} _{ — } _{F }of NAND gate **600** _{NAN2 }to be high, thereby accomplishing the desired result. In the opposite saturation case, when it is desired for the output bit of V_{EXEC} _{ — } _{F }to be a 0 (e.g., for the MSB in positive saturation or for all bits other than the MSB in negative saturation), then satx_{—}0 is asserted low and satx_{—}1 is asserted high. During this time, lsbz_mask is low because this is not a case of the EXTRACT instruction and, thus, there are two low signals input to OR gate **600** _{OR2}, forcing its output low and, thus, also forcing the output of NAND gate **600** _{NAN1 }high. This high is coupled with the high value of satx_{—}1 into NAND gate **600** _{NAN2 }produces a low value of V_{EXEC} _{ — } _{F}, thereby accomplishing the desired result.

Consider now the operation of circuit system **600** for the EXTRACT instruction where recall that the MSB(s) of V_{EXEC} _{ — } _{F }will include the sign bit while some the remaining LSB(s) of V_{EXEC} _{ — } _{F }will include rotated data. Toward this end, to provide a sign bit, mask_adj is low, and this low is input to NOR gate **600** _{NO1}, thereby causing it to output the complement of its other input. The other input is provided by the output of AND gate **600** _{AN1}, which for the EXTRACT instruction outputs sign_bitx because during that instruction data_gate is high, which therefore causes it to output sign_bit (i.e., the complement of sign_bitx), and sign bit is thus an input to OR gate **600** _{OR1}. The other input of OR gate **600** _{OR1 }receives a low value from the output of AND gate **600** _{AN2}, due also to the low value of mask_adj as input to AND gate **600** _{AN2}; thus, OR gate **600** _{OR1 }passes the value of sign_bit to an input of NAND gate **600** _{NAN1}, which receives at its other input a value of one because lsbz_mask is high (where its other input, satx_{—}0, is low for the EXTRACT instruction). Thus, NAND gate **600** _{NAN1 }outputs the complement of sign_bit and, thus, this output is again sign_bitx, which is connected as an input to NAND gate **600** _{NAN2}, which receives a high value at its other input because satx_{—}1 is also high for the EXTRACT instruction. Thus, NAND gate **600** _{NAN2 }complements the input of sign_bitx, thereby outputting the value of sign_bit for V_{EXEC} _{ — } _{F}. Also in connection with the EXTRACT instruction but to provide the rot_data as the output bit (i.e., as V_{EXEC} _{ — } _{F}), then mask_adj is asserted high, thereby causing AND gate **600** _{AN2 }to output rot_data to OR gate **600** _{OR1 }as one input, while the other input of OR gate **600** _{OR1 }will be low because it is provided by NOR gate **600** _{NO1}, which is forced to allow output by its high input value of mask_adj. Thus, OR gate **600** _{OR1 }passes rot_data as an input to NAND gate **600** _{NAN1}, which will receive a high at its other input from OR gate **600** _{OR2 }which receives a high input from lsbz_mask. Thus, NAND gate **600** _{NAN1 }outputs rot_datax, the complement of rot_data, but that complement is then inverted again by NAND gate **600** _{NAN2 }because its input satx_{—}1 is also high for the EXTRACT instruction. Thus, rot_data is output as V_{EXEC} _{ — } _{F}.

Consider now the operation of circuit system **600** for the EXTRACT instruction in the case where L>R in which case the LSB(s) will include a value of zero due to the output of LSB(s) zero mask block **510**. This output is achieved in circuit system **600** by setting the lsbz_mask bit equal to zero; again, since satx_{—}0 is held low for the EXTRACT instruction, then the zero value of lsbz_mask together with the low satx_{—}0 cause OR gate **600** _{OR2 }to output a low to NAND gate **600** _{NAN1}, thereby causing its output to be high. That his is ANDed with satx_{—}1, which for the EXTRACT instruction is high, thereby providing a low output of NAND gate **600** _{NAN2 }and, thus, providing the zero for the LSB(s) zero mask block **510**.

Completing the operation of circuit system **600**, the value of the data_gatex also may be provided as V_{EXEC} _{ — } _{F }in the same manner as described above with respect to outputting the sign_bitx bit. From the above, therefore, one skilled in the art will appreciate that this can be achieved by setting mask_adj to a low value and by providing sign_bitx as a high value.

**590** of **590** _{DL }because it implements dynamic logic, where the schematic is therefore repeated for a total of 32 bits to accommodate the 32-bit thermometer codes and to output the control bit code, CB[31:0], to multiplexer **502** for various instructions. In general, block **590** _{DL }receives dual rail input signal of the various mask signals and it provides dual rail outputs, mask and maskx, where those signals provide dual rail signals for the control bit code, CB[31:0]. Looking first to the devices and connections of block **590** _{DL}, and beginning toward the bottom left of the Figure, a node **610** _{ND1 }is connected to the drain of a p-channel transistor **610** _{P1}, and node **610** _{ND1 }is also connected to the drain of a p-channel transistor **610** _{P2 }the gate of a p-channel transistor **610** _{P3}, and an input of an inverter **610** _{INV1}. The sources of p-channel transistors **610** _{P1}, **610** _{P2}, and **610** _{P3 }are all connected to V_{DD}, and the gate of p-channel transistor **610** _{P1}, is connected to receive the clock signal, CLK. The gate of p-channel transistor **610** _{P2 }is connected to a node **610** _{ND2}, and node **610** _{ND2 }is also connected to the drain of p-channel transistor **610** _{P3}, the drain of a p-channel transistor **610** _{P4}, and to the input of an inverter **610** _{INV2}. The source of p-channel transistor **610** _{P4 }is connected to V_{DD }and its gate is connected to receive CLK.

Returning to node **610** _{ND1}, it is connected to various conditional discharge paths as follows. Between node **610** _{ND1 }and a node **612** _{ND1 }are connected, in series, the source/drain paths of three n-channel transistors **612** _{N1}, **612** _{N2}, and **612** _{N3}, having their respective gates connected to receive the shift_right_mask, setclr_mask, and setclr signals, the last of which is asserted during the evaluate phase of block **590** _{DL }when the SET or CLEAR instruction is executed. Between node **612** _{ND1 }and the source of n-channel transistor **612** _{N1 }is connected the source/drain path of an n-channel transistor **612** _{N4}, having it source connected to an shr signal, which is asserted during the evaluate phase of block **590** _{DL }when the SHIFT RIGHT instruction is executed. Also between nodes **610** _{ND1 }and **612** _{ND1 }are connected, in series, the source/drain paths of two n-channel transistors **612** _{N5 }and **612** _{N6}, having their respective gates connected to receive the shiftleft_mask and shl signals, the last of which is asserted during the evaluate phase of block **590** _{DL }when the SHIFT LEFT instruction is executed. Still further, between nodes **610** _{ND1 }and **612** _{ND1 }is connected the source/drain path of an n-channel transistor **612** _{N7}, having its gate connected to receive the fsh signal, which is asserted during the evaluate phase of block **590** _{DL }when the FUNNEL SHIFT instruction is executed.

Returning to node **610** _{ND2}, it is connected to various conditional discharge paths as follows. Between node **610** _{ND2 }and a node **614** _{ND1 }are connected, in series, the source/drain paths of two n-channel transistors **614** _{N1}, and **614** _{N2}, having their respective gates connected to receive the shiftright_maskx and shr_setclr signals, the last of which is asserted during the evaluate phase of block **590** _{DL }when any of the SHIFT RIGHT, SET, or CLEAR instructions is executed. Also between nodes **610** _{ND2 }and **614** _{ND1 }are connected the source/drain paths of two n-channel transistors **614** _{N3 }and **614** _{N4}, having their respective gates connected to receive the setclr_maskx and setclr signals, the last of which is asserted during the evaluate phase of block **590** _{DL }when either the SET or CLEAR instructions is executed. Also between nodes **610** _{ND2 }and **614** _{ND1 }are connected the source/drain paths of two n-channel transistors **614** _{N5 }and **614** _{N6}, having their respective gates connected to receive the shiftleft_maskx and shl signals, the latter of which is asserted during the evaluate phase of block **590** _{DL }when the SHIFT LEFT instruction is executed. Lastly and once more between nodes **610** _{ND2 }and **614** _{ND1 }is connected the source/drain path of an n-channel transistor **614** _{N7}, having its gate connected to receive the shrx_shlx_setclrx signal, which is asserted low when any one of the SHIFT RIGHT, SHIFT LEFT, SET, or CLEAR instructions is executed.

Finally, each inverter **610** _{INV1 }and **610** _{INV2 }has three transistors connected in association with the respective inverter. With respect to inverter **610** _{INV1}, its output is connected to the gate of a p-channel transistor **616** _{P1}, which has its source connected to V_{DD }and its drain connected to the input of inverter **610** _{INV1}. The output of inverter **610** _{INV1 }is also connected to the gate of an n-channel transistor **616** _{N1}, which has its source connected to ground and its drain connected to the source of an n-channel transistor **616** _{N2}. The gate of n-channel transistor **616** _{N2 }is connected to receive CLK, and the drain of n-channel transistor **616** _{N2 }is connected to the input of inverter **610** _{INV1}. With respect to inverter **610** _{INV2}, its output is connected to the gate of a p-channel transistor **618** _{P1}, which has its source connected to V_{DD }and its drain connected to the input of inverter **610** _{INV 2}. The output of inverter **610** _{INV2 }is also connected to the gate of an n-channel transistor **618** _{N1}, which has its source connected to ground and its drain connected to the source of an n-channel transistor **618** _{N2}. The gate of n-channel transistor **618** _{N2 }is connected to receive CLK, and the drain of n-channel transistor **618** _{N2 }is connected to the input of inverter **610** _{INV2}.

The operation of block **590** _{DL }is comparable in a logic sense to that described of the static logic implementation of block **590** and discussed above with respect to **590** _{DL}. In addition, the shiftleft_mask input signal is derived from the shiftright_mask input by swapping (i.e., inverting) the dual rail signals and downshifting by one bit. The positive rail of the dynamic logic of block **590** _{DL }(i.e., to the bottom left of

**600** _{DL }for implementing various of the signal selectivity provided by multiplexer **502**, saturation constant switch **582**, and AND gate **504** of _{EXEC} _{ — } _{F }and V_{EXEC} _{ — } _{FX}; thus, system **600** _{DL }in the preferred embodiment is duplicated for a total of 32 like-design circuits. Looking to the devices and connections in **700** _{IND1 }is connected to the drain of an n-channel transistor **700** _{N1}, which has its source connected to ground and its gate connected to receive a signal ext_sshl, which is asserted for execution of the SATURATING SHIFT LEFT instruction. Node **700** _{ND1 }is also connected to the source of an n-channel transistor **700** _{N2}, the source of an n-channel transistor **700** _{N3}, and the source of an n-channel transistor **700** _{N4}. The gate of n-channel transistor **700** _{N2 }is connected to receive a signal sat_{—}1, which is asserted during saturation if a one output is desired, and the drain of n-channel transistor **700** _{N2 }is connected to a node **700** _{ND2}. The drains of n-channel transistors **700** _{N3 }and **700** _{N4 }are connected to a node **700** _{ND3}, and their respective gates are connected to receive signals lsbz_maskx and sat_{—}0, where lsbz_maskx is only valid during an EXTRACT instruction and for other instructions is not enabling and where sat_{—}0 is disabling for the EXTRACT instruction and is asserted during saturation, for the SATURATING SHIFT LEFT instruction, if a zero output is desired. Returning to node **700** _{ND2}, between it and a node **700** _{ND4 }are connected, in parallel, the source/drain paths of two n-channel transistors **700** _{N5 }and **700** _{N6}, with their gates connected, respectively, to signals lsbz_mask and satx_{—}1. Node **700** _{ND4 }is also connected to the source/drain path of three n-channel transistors, **700** _{N7}, **700** _{N8}, and **700** _{N9}, having their respective gates connected to receive the signals mask_adj, sign_bit, and data_gate. The source of n-channel transistor **700** _{N7 }is connected to receive a bit of the active low data D_{LS }(i.e., data from data rotate unit **200**), while the sources of n-channel transistors **700** _{N8 }and **700** _{N9 }are connected to a node **700** _{ND5}. Between node **700** _{ND5 }and ground is connected the source/drain path of an n-channel transistor **700** _{N10}, which has its gate connected to receive a signal mask_adjx. Returning to node **700** _{ND3}, connected between it and a node **700** _{ND6}, in parallel are two n-channel transistors **700** _{N11 }and **700** _{N12}, having their respective gates connected to receive the signals satx_{—}0 and lsbz_mask. Node **700** _{ND6 }is also connected to the drain of three n-channel transistors **700** _{N13}, **700** _{N14}, and **700** _{N15}. N-channel transistor **700** _{N13 }has its gate connected to receive mask_adj and it source connected to receive an inverse of a bit from data rotate unit **200**, indicated in _{LS}, while n-channel transistors **700** _{N14 }and **700** _{N15 }have their respective gates connected to receive signals sign_bitx and data_gatex, both of which represent negative rail signals corresponding to the output of sign bit select block **506** and data gate block **530**. The sources of n-channel transistors **700** _{N14 }and **700** _{N15 }are connected to node **700** _{ND5}. Completing the bottom of **700** _{P1 }and **700** _{P2 }have their sources connected to V_{DD}, their gates connected to receive a clock signal, CLK, and their drains connected respectively to nodes **700** _{ND5 }and **700** _{ND6}.

Looking toward the top left of **700** _{ND2}, it is connected to the drains of two p-channel transistors **700** _{P3 }and **700** _{P4}, both having their sources connected to V_{DD}, and where the gate of p-channel transistor **700** _{P4 }is connected to the gate of an n-channel transistor **700** _{N16 }and to CLK, and the gate of p-channel transistor **700** _{P3 }is connected to node **700** _{ND3}. Node **700** _{ND3 }is comparably connected to the drains of two p-channel transistors **700** _{P5 }and **700** _{P6}, both having their sources connected to V_{DD}, and where the gate of p-channel transistor **700** _{P6 }is connected to the gate of an n-channel transistor **700** _{N17 }and to CLK, and the gate of p-channel transistor **700** _{P5 }is connected to node **700** _{ND2}. Continuing from node **700** _{ND2}, it is connected to the gate of a p-channel transistor **700** _{P7}, having its source connected to V_{DD }and its drain connected to both the output V_{EXEC} _{ — } _{F }and to the drain of an n-channel transistor **700** _{N18}, which has its gate connected to an enable signal and its source connected to the drain of an n-channel transistor **700** _{N19}, which has it gate connected to the gate of p-channel transistor **700** _{P7 }and its source is connected to ground. The output, V_{EXEC} _{ — } _{F}, is also connected to the gate of a p-channel transistor **700** _{P8}, having its source connected to V_{DD }and its drain connected to node **700** _{ND2}, which is also connected to the drain of n-channel transistor **700** _{N16}. The source of n-channel transistor **700** _{N16 }is connected to the drain of an n-channel transistor **700** _{N20}, which has its source connected to ground and its gate connected to the output, V_{EXEC} _{ — } _{F}.

Looking toward the top right of **700** _{ND3}, it is connected to the gate of a p-channel transistor **700** _{P9}, having its source connected to V_{DD }and its drain connected to both the complement output V_{EXEC} _{ — } _{FX }and to the drain of an n-channel transistor **700** _{N21}, which has its gate connected to the enable signal and its source connected to the drain of an n-channel transistor **700** _{N22}, which has it gate connected to the gate of p-channel transistor **700** _{P9 }and its source connected to ground. The complementary output, V_{EXEC} _{ — } _{FX}, is also connected to the gate of a p-channel transistor **700** _{P10}, having its source connected to V_{DD }and its drain connected to node **700** _{ND3}, which is also connected to the drain of n-channel transistor **700** _{N17}. The source of n-channel transistor **700** _{N17 }is connected to the drain of an n-channel transistor **700** _{N23}, which has its source connected to ground and its gate connected to the complementary output, V_{EXEC} _{ — } _{FX}.

The operation of dynamic logic circuit system **600** _{DL }should be appreciated by one skilled in the art given the comparable signals presented in it relative to the static approach set forth in **700** _{N7}, it receives mask_adj at its gate, which recall is one bit of the control bit code, CB[31:0], output by mask adjust block **590**. Thus, if mask_adj is asserted high, then n-channel transistor **700** _{N7 }is enabled and the rotated data bit, D_{LS}, is connected to node **700** _{ND4}. Further, for the EXTRACT instruction, satx_{—}1 is zeroed and lsbz_mask is one bit of the output of block **510**, thereby potentially enabling n-channel transistor **700** _{N5 }(while for all other instructions lsbz_mask is zeroed). Thus, in the cases when the one bit of lsbz_mask is high, the D_{LS }data at node **700** _{ND4 }data passes to node **700** _{ND2}. Further, assuming the enable signal is also asserted at this time (as is the case unless it is desired to place circuit **600** _{DL }in a tri-state setting), then the rotated data bit, D_{LS}, at node **700** _{ND5 }is inverted by the combination of p-channel transistor **700** _{P7 }and n-channel transistor **700** _{N19}, and that inverted data is therefore provided as the output V_{EXEC} _{ — } _{F}. At this same time and relative to the complementary output, V_{EXEC} _{ — } _{FX}, it provides the complement of DX_{LS}, because n-channel transistor **700** _{N13 }is also enabled, as is n-channel transistor **700** _{N12}, thereby connecting DX_{LS }to node **700** _{ND3}, and the data at that node is inverted by the combination of p-channel transistor **700** _{P9 }and n-channel transistor **700** _{N22}. Looking now to n-channel transistor **700** _{N10}, it receives at its gate mask_adjx, which is asserted high in the instance corresponding to an output of “0” for mask_adj, thereby enabling n-channel transistor **700** _{N10}. In response, either n-channel transistor **700** _{N8 }or n-channel transistor **700** _{N9 }may be enabled based on the state of its corresponding gate signal sign bit and data_gate (where only one of those can be high at a time based on the instruction being executed), with a complementary operation occurring for n-channel transistors **700** _{N14 }and **700** _{N15}, respectively. Thus, one of nodes **700** _{ND2 }or **700** _{ND3 }is discharged through the enabled transistors, and that discharged state is inverted and output at one of V_{EXEC} _{ — } _{F }or V_{EXEC} _{ — } _{FX}, with the complementary state being output at the other of those outputs. Looking also at the operation of circuit **600** _{DL }for the SATURATING SHIFT LEFT instruction, if there is saturation and a 1 output is desired then sat_{—}1 is asserted high while satx_{—}0 is low, whereas if there is saturation and a 0 output is desired then sat_{—}0 is asserted high while satx_{—}1 is low. A high value of sat_{—}1 enables n-channel transistor **700** _{N2 }thereby discharging node **700** _{ND2 }and providing V_{EXEC} _{ — } _{F }high, whereas a high value of sat_{—}0 enables n-channel transistor **700** _{N4 }thereby discharging node **700** _{ND3 }and providing V_{EXEC} _{ — } _{FX }high (and, at the same time, V_{EXEC} _{ — } _{F }is low).

From the above, it may be appreciated that the above embodiments provide for various improvements over the prior art. As one benefit, the preferred embodiments provide an electronic processor operable in response to an instruction to efficiently shift a data argument in response to respective left and right shift count indicators, without the need to separately shift the entire data argument fully in response to a first shift count followed by then separately shifting the entire data argument fully in response to a second shift count. As another benefit, the electronic processor may be further modified to accommodate any one or more of numerous data manipulation instructions. As another benefit, various implementation details for the preferred embodiments have been provided, including approaches using either static or dynamic logic. As still another benefit, while the present embodiments have been described in detail, various substitutions, modifications or alterations could be made to the descriptions set forth above without departing from the inventive scope. For example, while a 32-bit data argument has been shown, the inventive teachings may be applied to other sizes of data argument. As another example, while the ROTATE STAGEs have been shown in a particular order, that order is only to simplify an understanding of the operation with respect to the order of the bit positions in the shift arguments; however, since certain shifts are independent of others, some STAGEs may be placed in different orders. Thus, from these many benefits as well as the preceding details, one skilled in the art should further appreciate the inventive scope, as is defined by the following claims.

Patent Citations

Cited Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US4396994 | Dec 31, 1980 | Aug 2, 1983 | Bell Telephone Laboratories, Incorporated | Data shifting and rotating apparatus |

US4653019 | Apr 19, 1984 | Mar 24, 1987 | Concurrent Computer Corporation | High speed barrel shifter |

US5073864 * | Feb 12, 1990 | Dec 17, 1991 | Davin Computer Corporation | Parallel string processor and method for a minicomputer |

US5751614 * | Feb 29, 1996 | May 12, 1998 | Exponential Technology, Inc. | Sign-extension merge/mask, rotate/shift, and boolean operations executed in a vectored mux on an ALU |

US5781457 * | May 14, 1996 | Jul 14, 1998 | Exponential Technology, Inc. | Merge/mask, rotate/shift, and boolean operations from two instruction sets executed in a vectored mux on a dual-ALU |

Non-Patent Citations

Reference | ||
---|---|---|

1 | "Fast Masked-Based Bit Extraction and Justification" Research Disclosure, Kenneth Mason Publications, No. 439, No. 1, 2000, p. 2028. | |

2 | D.D. Gajski et al "High-Speed Masking Rotator" Digital Processes, Georgi Publishing Company, vol. 4, 1978, pp. 67-81. |

Referenced by

Citing Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US8041755 * | Jun 8, 2007 | Oct 18, 2011 | Apple Inc. | Fast static rotator/shifter with non two's complemented decode and fast mask generation |

US8704689 * | Jan 25, 2013 | Apr 22, 2014 | Texas Instruments Incorporated | Method of processing data samples and circuits therefor |

US8768989 | Jun 20, 2011 | Jul 1, 2014 | Apple Inc. | Funnel shifter implementation |

US8972469 | Jun 30, 2011 | Mar 3, 2015 | Apple Inc. | Multi-mode combined rotator |

US9015216 | Sep 14, 2011 | Apr 21, 2015 | Apple Inc. | Fast static rotator/shifter with non two's complemented decode and fast mask generation |

US20050273484 * | Jun 6, 2005 | Dec 8, 2005 | Jia-Horng Shieh | Method and apparatus for processing multiple decomposed data for calculating key equation polynomials in decoding error correction code |

US20080307204 * | Jun 8, 2007 | Dec 11, 2008 | Honkai Tam | Fast Static Rotator/Shifter with Non Two's Complemented Decode and Fast Mask Generation |

US20130194112 * | Jan 25, 2013 | Aug 1, 2013 | Texas Instruments Incorporated | Method of processing data samples and circuits therefor |

Classifications

U.S. Classification | 708/209, 712/E09.034 |

International Classification | G06F7/00, G06F9/00, G06F9/315, G06F9/305, G06F7/76, G06F15/00 |

Cooperative Classification | G06F9/30032 |

European Classification | G06F9/30A1M |

Legal Events

Date | Code | Event | Description |
---|---|---|---|

Dec 20, 2002 | AS | Assignment | Owner name: TEXAS INSTRUMENTS INCORPORATED, TEXAS Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BOSSHART, PATRICK W.;REEL/FRAME:013642/0919 Effective date: 20021220 |

Jan 27, 2012 | FPAY | Fee payment | Year of fee payment: 4 |

Jan 25, 2016 | FPAY | Fee payment | Year of fee payment: 8 |

Rotate